ANALYSIS OF DIAGNOSTIC TESTING OF SUSTAINED CASING PRESSURE IN WELLS

A Dissertation Submitted to the Graduate Faculty of the Louisiana State University and Agriculture and Mechanical College in partial fulfillment of the Requirements for the degree of Doctor of Philosophy

in The Department of Petroleum Engineering

by Rong Xu B.S., University of Petroleum (East China), 1993 M.S., University of Petroleum (Beijing), 1996 December 2002

ACKNOWLEGEMENTS
The author extends her gratitude to Dr. Andrew K. Wojtanowicz, under whose precious guidance, advice and encouragement, this work was accomplished successfully. His timely suggestions have always been valuable and helpful during the course of this research. Deep appreciation is extended to Dr. Zaki Bassiouni, Dr. John Rogers Smith, Dr. Christopher White, and Dr. Mehdy Sbbaghian, for their support and constructive advice. This research was financed through funds made available by the Mineral Management Service of the United States Department of the Interior with supplementary support from the Craft and Hawkins Department of Petroleum Engineering at Louisiana State University. Without their support, this work would not have been possible. In addition, thanks are extend to all other faculty members and students who have offered help and been kind to the author and made the past four years enjoyable and memorable.

ii

TABLE OF CONTENTS
ACKNOWLEGEMENTS..................................................................................................ii ABSTRACT....................................................................................................................... v CHAPTER 1. INTRODUCTION ...................................................................................... 1 1.1 Field Data Analysis..................................................................................................2 1.1.1 Statistical Analysis of SCP Data Bank ............................................................. 2 1.1.2 SCP Typical Patterns ........................................................................................4 1.2 Current Procedures for the SCP Diagnostic Test.....................................................6 CHAPTER 2. MECHANISMS INVOVLED IN SCP TESTS - CRITICAL REVIEW..10 2.1 Late Gas Migration in Cement...............................................................................14 2.2 Effect of Bubble Generation on Bubble Migration ...............................................15 2.3 Gas Migration In Stagnant Mud ............................................................................16 2.3.1 Flow-Pattern Identification for Two-Phase Flow in Annuli ........................... 19 2.3.2 Gas Slip Velocity ............................................................................................25 2.3.3 Gas Distribution in Mud Column above Cement ...........................................33 2.3.4 Frictional Pressure Loss ..................................................................................38 2.4 Flow Through Chokes ...........................................................................................41 2.4.1 Single-Phase Gas Flow ................................................................................... 43 2.4.2 Multiphase Flow .............................................................................................46 CHAPTER 3. CASING PRESSURE BUILDUP IN WELLS WITH GAS-FREE LIQUID COLUMN ABOVE CEMENT .........................................................................51 3.1 Model Description .................................................................................................51 3.2 Parameters Affecting Pressure Buildup.................................................................52 3.3 Model Validation with Data from Wells ...............................................................55 3.3.1 Analysis of SCP in Well 23 ............................................................................ 56 3.3.2 Analysis of SCP in Well 24 ............................................................................ 57 CHAPTER 4. MATHEMATICAL MODEL OF SCP TEST.......................................... 59 4.1 Gas Flow in Cement ..............................................................................................59 4.2 Gas Migration in Mud Column..............................................................................60 4.2.1 Numerical Procedure ......................................................................................65 4.2.2 Boundary Conditions ...................................................................................... 72 4.2.3 Solution Algorithm ......................................................................................... 75 4.3 Coupling Gas Flows in Liquid Column and Cement.............................................77 CHAPTER 5. THEORETICAL ANALYSIS OF SCP TESTING .................................. 81 5.1 Understanding of SCP Typical Patterns ................................................................81 5.1.1 Bleed-Down Patterns ...................................................................................... 81 5.1.2 Buildup Patterns..............................................................................................83 5.2 Understanding of SCP Testing Scenarios ..............................................................86 5.2.1 Thin Mud ........................................................................................................86 5.2.2 Thick Mud.......................................................................................................88

iii

5.2.3 Slightly Compressible Mud vs. Incompressible Mud.....................................90 5.3 Sensitivity Analysis ...............................................................................................90 5.3.1 Approach.........................................................................................................91 5.3.2 Methods of Sensitivity Analysis of SCP Testing............................................ 94 5.3.3 Analysis of Bleed-Down Pressure ..................................................................98 5.3.4 Analysis for Buildup Pressure ........................................................................99 5.3.5 Discussion of Sensitivity Analysis................................................................102 5.4 Field Validation of SCP Test Model....................................................................103 5.4.1 Validation with Data from Well 19...............................................................103 5.4.2 Validation with Data from Well 25...............................................................105 CHAPTER 6. SUMMARY AND CONCLUSIONS .....................................................107 REFERENCES ..............................................................................................................111 APPENDIX A. DATA BANK AND SOFTWARE ......................................................117 APPENDIX B. SCP BUILDUP IN WELLS WITH GAS-FREE LIQUID COLUMN ABOVE CEMENT ........................................................................................................118 APPENDIX C. SOLUTION OF THE DIFFERENTIAL EQUATION FOR THE LINEAR GAS FLOW....................................................................................................122 APPENDIX D. MANUAL OF SOFTWARE: SCPTESTMODEL............................... 133 APPENDIX E. ABSTRACT OF PAPER SPE 67194 ................................................... 139 APPENDIX F. PRESSURE HISTORY FOR WELL 19 & 25 ..................................... 141 VITA .............................................................................................................................142

iv

ABSTRACT
Over 8,000 wells in the Gulf of Mexico exhibit sustained casing pressure (SCP). SCP is defined as any measurable casing pressure that rebuilds after being bled down, attributable to cause(s) other than artificially applied pressures or temperature fluctuations in the well. The Minerals Management Service (MMS) regulations consider SCP hazardous and, in principle, require its elimination. In some cases the agency may allow continuing production at a well with SCP by granting a temporary departure permit. The departure permits are based on diagnostic tests involving pressure bleed-down through a 0.5-inch needle valve followed by closing the valve and recording pressure buildup for 24 hours. Presently, analysis of testing data is mostly qualitative and limited to arbitrary criteria. This work provides theory, mathematical models and software needed for qualitative analysis of SCP tests. SCP occurs due to the loss of wells external integrity causing gas inflow from a high-pressure formation into the wells annulus. Then, the gas migrates upward through a leaking cement sheath, percolates through the mud column and accumulates above the liquid level inside the gas cap. The study identified two scenarios of gas flow in the liquid column: rapid percolation through low-viscosity Newtonian fluid; and, slow ascendance of gas bubble swarms in viscous, non-Newtonian mud. The two scenarios have been mathematically modeled and theoretically studied. The first model assumes rapid percolation and ignores gas entrainment in the liquid column. Simulation showed that early pressure buildup was controlled by mud compressibility, annular conductivity, and gas cap volume while formation pressure

controlled the late pressure buildup. Mathematical simulations matched pressure buildups recorded in two wells, showing that the model had physical merit. The second mathematical model fully describes gas migration by coupling the variable rate gas flow in cement with the two-phase flow in liquid column. The model was used to study typical patterns of bleed-down and buildup from SCP diagnostic tests. It showed that analysis of pressure bleed-down gives properties of gas-liquid mixture above the cement, while a sufficiently long pressure buildup may give values of the annular conductivity, the depth and pressure of the gas-source formation.

vi

CHAPTER 1. INTRODUCTION
Sustained Casing Pressure (SCP) is seen in over 11,000 casing strings in over 8000 wells on the Outer Continental Shelf (OCS) (Bourgoyne, Scott, and Manowski, refer to Chapter 1). A typical well completion places cement to seal off the interior of various casing strings from the subsurface formations (Fig.1.1). Ideally in the well of Fig.1.1, only production tubing should be pressured. Gauges on all of the casing strings should read zero after a small volume of fluid caused by thermal expansion effects has been bled. If the casing pressure returns when the needle valve is closed, then the casing is said to exhibit SCP. The Minerals Management Service (MMS) is concerned about wells on the OCS that exhibit significant sustained casing pressure because of its responsibility for worker safety and environmental protection as mandated by Congress.

Fig.1. 1 Simplified well schematic In 1988, a regulation (a part of that is now 30 CFR 250.517) required that all annuli be monitored for SCP and that every occurrence of SCP be reported immediately to the District Supervisor. Implementation of this regulation imposed a heavy regulatory burden 1

on both the lessee and MMS. As a result, discussions were initiated between MMS and the Offshore Operators Committee (OOC). After some discussion the OOC commenced a study of SCP (Bourgoyne, Scott, and Manowski, 1998, refer to Chapter 3). 1.1 Field Data Analysis A SCP data bank was developed from field data (Appendix A). The field data are casing pressure records provided by various operators from 26 wells and are contained in Microsoft Excel (.xls) files and corresponding well numbers are listed in Table 1.1. Each file has a worksheet of raw data. Usually, charts include only the casing strings that have SCP problems, and chart names are the outer diameters of those strings. In some cases, if the string has more than one cycle of pressure buildup, each period has a separate chart. 1.1.1 Statistical Analysis of SCP Data Bank SCP Occurrence by Casing String. Among those 26 wells in the data bank, 22 wells, 85% of the total, have SCP problems. SCP occurrence by casing type was concluded in Table 1.1. As indicated by the table, the following trends may be observed: About 21% of the casing strings exhibiting SCP are production casing. About 45% of the casing strings exhibiting SCP are intermediate casing strings. About 14% of the casing strings exhibiting SCP are surface casing strings. About 11 % of the casing strings exhibiting SCP are conductor casing strings.

The statistical analysis shows the trend similar to that reported by MMS (Fig.2.1), with exception of the intermediate casing strings (10%). SCP Magnitude by Casing String. Shown in Fig.1.2 is a cumulative frequency plot of the occurrence of the SCP magnitude in unit of psi for the various types of casing strings. Among casings with SCP problem, about 50 percent of the production casings

and 35 percent of the intermediate casings have SCP of less than 1000 psi, comparing that about 80 percent of the production casings and intermediate casings have SCP in the GOM-MMS database. For the other casing strings, more than 90 percent of the strings have SCP of less than 500 psi. Same conclusion of other casing strings was also drawn from the statistical analysis in the GOM-MMS database. Table 1. 1 SCP Occurence in Various Casing Strings
Production Casing Intermediate Casing Surface Casing Well # File Name 6 5/8" 7" 7 5/8" 8 5/8" 9 5/8" 10 3/4" 11 3/4" 13 3/8" 16" 1 MUA1 NA NA Y 2 MUA2 Y N Y 3 MUA3 Y Y Y 4 MUA4 Y Y N 5 MUA5 Y Y N 6 MUA6 NA NA N 7 MUA7 N N N 8 MUA8 Y Y Y 9 MUA9 Y Y Y 10 MUA10 Y Y Y 11 MUA11 N N Y 12 MUA12 Y Y Y 13 MUA13 N N N 14 MUA15 N Y N 15 MUA16 N N N 16 APTA19 NA Y NA 17 APTA30 NA NA NA 18 APTA31 NA Y NA 19 APTL9 NA Y NA 20 BPTB6 NA Y NA 21 PTCA25C NA Y NA 22 PTCA7D NA NA Y 23 B7 N Y N 24 HIA1 N Y 25 HIA2 N Y 26 HIA3 N Y Total 0 8 0 0 8 9 1 8 0 PSCP % 21 45 24 Conductor 16" 20" N Y N N N N N N Y N Y N N N N NA Y NA NA NA NA NA N N N N 2 2 15

Y- SCP problem;

N- no SCP problem;

NA - data not available

100% 80% 60% 40% 20% 0% Production Casing Intermediate Casing < 1000 psi < 1000 psi

< 500 psi < 500 psi

Surface Casing

Conductor Casing

Fig.1. 2 Frequency of SCP for different casings in 26 wells

1.1.2 SCP Typical Patterns From the field data, five typical response patterns were concluded, including two SCP bleed-down patterns and three SCP buildup patterns. Instant Bleed-Down Patterns. This pattern is common when using bleed and lubricate remediation method. During bleed-down, the needle valve has been opened widely to bleed small amount of gas and liquid from the casing annulus for a very short time. As shown in Fig.1.3, casing pressure dropped from 680 psi to 40 psi in just 14 seconds and 30 gal of 17.2 ppg mud is removed.

800

Casing Pressure (psi)

600 Removed: 30 gal 17.2# fluid 400

200

14 Seconds

0 0 5 Time (sec) 10 15

Fig.1. 3 10 casing pressure of well 25 during bleed-down

Long Bleed-Down Pattern. On the other hand, some operators control the opening of needle valve to minimize the removal of fluid from the casing annulus. The duration of bleed-down is prolonged by the manipulation. For the intermediate casing in well 19 (Fig.1.4), casing pressure decreased non-linearly from 1340 psi to 749 psi in 12 minutes.

1400

Casing Pressure (psi)

1200

1000 12 minutes 800 Non-linear

600 0 2 4 6 8 10 12 14 Time (min)

Fig.1. 4 10 casing pressure of well 19 during bleed-down Normal Buildup Pattern. Shown in Fig.1.5 is the normal behavior of casing pressure buildup in a well with SCP problem. The casing pressure rises quickly after the bleed-down (early time behavior) and ultimately stabilizes at a certain level (late-time behavior). The transition is gradual pressure increase. The stabilized casing pressure is determined by mud weight and gas-source formation pressure. Transient time is controlled by the magnitude of gas migration in the cement and mud column.
1500 1400 Casing Pressure (psi) 1300 1200 1100 1000 900 800 0 5 10 15 20 25 Time (hours)

Transition

Late Time

Fig.1. 5 10 casing pressure of Well 18 5

S-Shape Buildup Pattern. As shown in Fig 1.6, there is no obvious increase of casing pressure until the first swarm of gas reaches at casing head. Then casing pressure increase gradually. Finally, the pressure stabilizes at certain level to finish one cycle of buildup.

140 120 Casing Pressure (psi) 100

60 40 20 0 0

Early Time

80

Transition

Late Time

10

15

20

25

Time (hours)

Fig.1. 6 13 3/8 casing pressure of Well 22 Incomplete Buildup Pattern. Shown is Fig.1.7 is an incomplete SCP response. After the bleed-down, casing pressure continually increases. No late-time stabilization is apparent in the testing interval (usually 24 hours). Compared with normal pattern, casing pressure increase at early time is relatively low. In this casing, the pressure increases about 24% in first two hours. 1.2 Current Procedures for the SCP Diagnostic Test The following concept of departure has been based on understanding that small and intermittent pressure induces least risk. However, technical criteria, which are based on the ratio of casing pressure to its strength and the ability to bleed to the zero pressure, are arbitrary to some degree. 6

1400

Casing Pressure (psi)

Early Time

1200

Transition

1000

800 2 hours 600 0 5 10 15 20 25 Time (Hours)

Fig.1. 7 10 3/4 casing pressure of Well 19 MMS has developed guidelines under which the offshore operator could selfapprove a departure for 30 CFR 250.517. Departure approval is automatic as long as the SCP is less than 20% of the minimum internal yield pressure and will be bled down to zero through a 0.5-in. needle valve in less than 24 hours. Diagnostic testing of all casing strings in the well is required if SCP is seen on any casing string. Records of each diagnostic test must be maintained for each casing annulus with SCP. The diagnostic tests must be repeated whenever the pressure is observed to increase (above the value that triggered the previous test) by more than 100 psi on the conductor or surface casing or 200 psi on the intermediate or production casing. Well operations such as acid stimulation, shifting of sliding sleeves, and replacement of gas lift valves also require the diagnostic tests to be repeated. If at any time the casing pressure is observed to exceed 20% of the minimum internal yield pressure of the affected casing, or if the diagnostic test shows that the casing will not bleed to zero pressure through a 0.5in. needle valve over a 24 hour period, the operator is expected to repair the well under regulations 30 CFR 250.517. 7

The recent report on the SCP problem (OTC 11029, Bourgoyne et al., 1999) shows the technical complexity of the SCP mechanism and provides recommendations for changing the criteria used in the SCP risk evaluation. It suggests that the flow rates of gas and liquid causing the SCP should be included in the record. Also, the well should be regularly shut in and tested for casing pressure buildup behavior. Recently, MMS has proposed a modified procedure for diagnostic testing (MMS Draft NTL, January 2000). Under this guideline, operators must address all casing pressure diagnostics and departures on a whole well basis. This means that when any annulus on a well needs a diagnostics test, operators must diagnose all casings with SCP at the same time, unless Technical Assessment and Operation Support (TAOS) Section specifically directs otherwise. During a diagnostic test, operators must record all initial pressure and both bleed-down and buildup pressure graphically or tabularly in no greater than 1-hour increments for each casing annulus in the wellbore. Operators must bleeddown and build up separately. Also operators must record the rate of buildup of each annulus for the 24-hour period immediately following the bleed-down. If fluid is recovered during bleed-down, operators must record the type and amount. Operators should conduct bleed-down to minimize the removal of liquid from the annulus. For subsea wells, where only production annuli can be monitored, operators must conduct diagnostics as indicated, except that results for adjacent annuli will be restricted to monitoring tubing pressure response. Currently, pressure testing analysis is successful in identifying wellhead, tubing and production casing leaks. Those leaks can also be remedied by conventional workover operations. But for outer strings, no theory-supported casing pressure testing procedure

has been developed. Moveover, although MMS requires the casing pressure to be monitored in SCP diagnostic tests including bleed-down and buildup period, a practical analysis method for those data has not been developed. This research developed mathematical models for testing and identification of the flow mechanism for external leaks leading to SCP. To reach that, the focus was on: Identifying controlling parameters of SCP. Determining the possible flow patterns of gas flow in the annulus with stagnant nonNewtonian fluid. Developing mathematical models to provide theoretical support for diagnostic testing of wells with SCP. Using the model to obtain understanding of typical SCP patterns to analyze the diagnostic testing data and to assess the severity of the problem. Part of the research has been presented as SPE paper 67194 at 2001 Production and Operation Symposium in Oklahoma City (Appendix E).

CHAPTER 2. MECHANISMS INVOVLED IN SCP TESTS - CRITICAL REVIEW

The most significant cause of SCP in the outer casing strings, outside of the production casing, is gas migration through a poor cement bond. From the database in MMS, 50% of the casing strings exhibiting SCP are those strings (Fig.2.1 after Bourgoyne, Scott, and Manowski, 1998, refer to Chapter 3). During primary cementing, gas can invade the cement and form channels. Even after the cement sets, pressure (Jackson and Murphey, 1990) and temperature changes (Goodwin and Crook, 1993) caused by completion and production operations may contribute to development of cracks and micro-annuli in the cement sheath. Those channels and cracks provide paths for gas to migrate and accumulate at the wellhead, causing SCP.
% ALL CASINGS WITH SCP

60% (11,498 CASINGS WITH SCP IN 8122 WELLS ARE INCLUDED IN THIS GRAPH) 50% 40% 30% 20% 10% 0% PROD INTER SURF COND STRUCT

Fig.2. 1 SCP in combined database by casing string There are two possible configurations in the annulus such as cemented to the surface or a mud column above the cement (Fig.2.2). In annuli cemented to the surface, gas migration has been considered a flow through porous media having some permeability (Somei, 1999, refer to Chapter 4). After bleed-off, the casing pressure increase is analogous to the pressure-buildup (Fig.2.3). The buildup behavior is controlled by the cement properties such as permeability and porosity, as well as formation pressure. 10

Fig.2. 2 Two configurations of casing annuli. (A) Annulus cemented to the surface. (B) Annulus with a mud column above cement
Bleed off Bleed off

Pressure

Time

Fig.2. 3 Conceptual behavior of SCP Alternative mechanics are gas flowing through matrix channels or micro-annuli. Matrix channels exist in the cement itself. The micro-annulus is the path between cement and casing or cement and formation. If wells have a mud column above the cement (Fig.2.2B), gas flow in cement is affected both by formation and cement properties, and rheology of mud, since mud properties controls the gas flow rate at the interface between cement and mud column. 11

Gas flow in cement can be described is linear (vs. radial) gas flow with changing rate at cement top and constant pressure at the gas-source formation. When gas is released from the interface, flow mechanics changes. Porous cement can be considered as a bundle of orifices. If gas flow rate is high and mud gel strengh is low, small bubbles continuously form, grow and drift from the orifice (Fig.2.4A). On the other extreme case, mud gel-strength could so high that it acts as a rigid solid. A piston of gas may form and push up the whole mud column(Fig.2.4B). In SCP diagnostic tests, the fact that gas is bled from the casing annulus suggest that gas bubbles break the liquid gel and can migrate. Therefore, first mechnism is considered in in this study, i.e., gas bubbles constantly form and flow from the interface.

Fig.2. 4 Bubble generation at the interface. (A) Small bubbles form in slightly gelled Mud. (B) Gas slug forms in extremely gelled mud In mud column, the flow pattern depends on the operational variables (flow rate), phyiscal properties of the mud, and geometrical variables of the system (Cheremisinoff, 1986). When gas is introduced at low flow rate into a large vertical column of liquid, the gas phase is distributed into bubbles. If the bubbles are very small, they behave as rigid spheres rising vertically in rectilinear motion (Sangani and Acrivos, 1983). However, above a critical size (about 0.3 cm for air-water at low pressure) the bubbles begin to 12

deform, and upward motion is more erratic. The bubbles randomly collide and coalesce, forming Taylor bubbles. With an increase in gas flow rate, at low liquid rates, the bubble density increases and dispersed bubbles become so closely packed that many collisions occur, a transition to slug flow. Because the gas flow rate is fairly low in SCP buildup, the most likely flow patterns will be bubble and slug flow. In a usually stagnant mud column, if gas bubbles are intermittently released at the interface such as gas kicks, the gas velocity is solely the slip velocity driven by density diffference between gas and liquid. But if gas bubbles are continously generated at the interface, gas bubbles rise faster than single bubbles or growing bubbles which are not being continuously generated. An extra increment, propotional to the superfacial gas velocity must be added to the basic rising velocity. (Nicklin et al, 1962). In the case of SCP buildup, gas is assumed to flow continuously at the mud-cement interface until all pressures in the system stabilize. Therefore, a two-phase flow model instead of gas slip model is used. Gas either accumulates at the top portion of the annulus during buildup or discharges from casing head in bleed-down. The volume of this wellhead gas is affected by the gas releasing rate from cement top and gas expansion in the rest of the annulus. Those, in turn, depend on casing, mud and gas compressibility. In bleed-down, gas bleeding rate is an additional controlling factor. When the needle valve is closed, pressure change at wellhead is related to cummulative gas inflow by the real gas law due to the high gas concerntration in this part. While the valve is opened, surface casing pressure is determined by flow through the choke as well as mud, cement and formation properties. Most outer casing strings exhibiting SCP have mud columns above the cement in the annuli. The model of gas migrating through the cement and mud is more relevant for these

13

strings. Mechanisms that contribute to SCP in outer string include late gas migration in cement, bubble generation at the cement-mud interface, two-phase flow in mud column and gas accumulation/vacating at well head. Literature studies involving those mechanisms are discussed in following sections. 2.1 Late Gas Migration in Cement Gas migration is classified into two distinct groups primary and secondary. The former is related to aspects of the actual cementing operation i. e. slurry characteristics, displacement mechanics, hydrostatic pressure (Rae et al, 1989). Secondary or late gas migration, has little to do with the cementing operation. It may be caused by mechanical and thermal stresses which damage the integrity of hydraulic bond or the cement material and lead to gas leakage (Jackson et al; 1990, Goodwin et al; 1993). Maurice et al. (2000) proposed that cement shrinkage may cause long-term leakage. A circumferential fracture is propagated upward by the slow accumulation of gas under pressure behind the casing. Cement system with high water/cement ratio (low density, extended system) may be leakprone. These cements can exhibit fairly high innate permeability (0.5 5md), even when set (Sykes et al, 1987). It is possible, therefore, for gas to flow within the matrix of such cements and to reach the surface, resulting in SCP. Few researchers have related late gas migration to SCP testing. Somei (1999) developed a model of late gas migration through the annulus cemented to the surface. He considered the gas migration as a flow through permeable porous media. With assumptions of constant formation pressure, zero flow rate at wellhead, and initially steady-state flow, he derived an analytical solution similar to gas well testing solutions. He analyzed effects of cement porosity, temperature, and gas gravity on SCP buildup and concluded that low

14

porosity, low temperature and low gas gravity enhanced SCP buildup. However, he did not verify his solution with experimental or field data. Moreover, the model is limited because most outer casing is not cemented to the surface. 2.2 Effect of Bubble Generation on Bubble Migration Ebert (1980) studied bubble formation in a CMC-H2O (structurally viscous) solution. He concluded that bubble formation mechanism is influenced by non-Newtonian behavior only at a shear rate of g& 70 s -1 . Bubbles generated in viscoelastic liquid usually exhibit an elongated shape (Boehme 1982). Casariego (1987) reported his theoretical investigation on the initial bubble size when it was released in Bingham plastic liquids. The volume of bubble in each stage was determined by an iterative method. For the range of gas flow rate and orifices used in his experiments, the size of the bubbles was not strongly affected by the viscosity of the fluid. When the gas flow rate is lower than 3 10-6 scf/s, bigger bubbles form in the glycerine. The gel strength behavior of the mud system and its effect on the bubble rising velocity has been studied by Santos et al (1997). Quiescent time is the interval between mud circulation stopping and the bubble is release. The gel strength grows with quiescent time. They concluded that quiescent time does not affect the air migration rate (Table 2.1). Table 2. 1 Effect of Quiescent Time on Migration Rate in Bentonitic Mud Quiescent Time (s) 0 600 1200 Migration Rate (m/s) 0.321 0.316 0.320 Gel Strength (Pa) 3.352 6.224 6.224 From those literatures, it was concluded that the motion of gas bubbles in structurally viscous CMC-H2O solution is nearly independent of the type of generation. While in viscoelastic liquids, the generation of bubbles is a prerequisite for ascent. Further studies on two-phase flow show that once the bubble is generated and released from the interface, 15

its rising velocity depends on neither liquid gel strength nor its size. In this study, the effect of bubble generation was not considered and gas rising velocity was only affected by rheology of liquid and annular geometry. 2.3 Gas Migration In Stagnant Mud Slip velocity can describe gas migration in stagnant mud. A single bubble is released at the entry. It is a suitable model for gas kicks. After detection of kick, bottom hole pressure is maintained slightly higher than the formation pore pressure to prevent more gas flowing from the formation. Considering gas slip in mud, Mathews (1980) used a hydrostatic model to derive the surface pressure rise rate in gas-kick control. The surface pressure is related to the mean gas pressure by p s = p m - 0.052 r g Lg 2 - 0.052 r m D (2. 1)

where p s = surface casing pressure, psig p m = mean gas pressure (located at midpoint of gas slug), psig
r g = average gas density
D = depth to top of gas kick, feet

L g = length of gas slug, feet

His assumptions included incompressible mud, constant temperature profile and relatively constant gas compressibility factor (Z). With those assumptions, he further concluded that gas pressure inside the slug would remain essentially constant at the formation pore pressure as it migrates up the wellbore. Therefore, The surface pressure rise rate is given by 16

dp s dD = -0.052 r m = 0.052 r m v s dt dt

(2. 2)

where v s = average gas slip velocity, ft/s Johnson and Tarvin (1993) showed a more comprehensive shut-in wellbore model:
C g V g r m gv s - q e dp s = dt C g V g + C wVw + C mVm

(2. 3)

where, C g and V g are the influx (gas) compressibility and volume respectively, C w and V w the wellbore compressibility and volume and C m and Vm the mud compressibility and volume. q e is the fluid loss rate from the wellbore. All those models assume that gas pressure keep constant when gas bubbles rise through the mud. In fact, when gas bubbles migrate up, the pressure inside bubbles will always be equal to the pressure of mud around them. As gas accumulates at the top, gas pressure will decrease and bubbles will expand, showing that the assumption does not reflect the reality. Moreover, those techniques ignored the effect of continual inflow of gas. During SCP tests, we have no control at mud column bottom. As long as there is a pressure difference at mud-cement interface, gas will continuously flow through it and the flow rate will affect gas rising velocity. Therefore, the two-phase flow approach is more proper for describing gas migration in mud column in SCP. This conclusion is also proved by Zuber and Finadlay (1965). Time- and space-averaged velocities for gas (Zuber and Finadlay 1965) are given as

v g = C0 vm + vs
where, C 0 is the distribution factor and v s is called gas slip velocity.

(2. 4)

17

Griffith and Wallis (1959) applied the theory of continuity to show that the homogeneous velocity between the gas bubbles, v m , must be:
v m = (q g + q L ) / A = v Sg + v SL

(2. 5)

where, q g and q L are the respective volume flow rates of gas and liquid. And v Sg and v SL are the respective superficial velocities of gas and liquid. When q g = q L = 0 , the Eq.2.4 refer to a bubble rising in still liquid. Thus, v g = v s . When q L = 0 , it is the case that bubbling gas flows through stagnant liquid. So,

v g = C0

qg A

+ vs .

(2. 6)

Thus, even though the net flow of liquid is zero, the gas rises at a velocity greater than its slip velocity (Nicklin, 1961), which is the flow situation happened in SCP. The distribution factor, C 0 , is expected to be in the range 0.9< C 0 <2.0. It represents the enhancement of gas velocity that occurs if the gas is concentrated towards the center of the flowing region where the velocity is the greatest. Therefore, selection of its value depends on flow patterns. Zuber and Finadlay (1965) have studied the variation of C 0 for various assumed velocity and void profile. Typically, C 0 = 1.2 for bubble and slug flow pattern. While for annulus flow pattern, C 0 = 1.0 . By use of a drift-flux concept, Hasan and Kabir (1988) reported that
1.2 if c0 = 2.0 if d < 0.12m or if d > 0.12m and if v SL > 0.02 m v SL < 0.02 m s s (2. 7)

for bubble flow. While for slug flow, C 0 = 1.2 .

18

Johnson and Cooper (1993) reported experiments made in a 200 mm (7.8) pipe and a 200 89 mm (7.8 3.25) eccentric annulus with non-Newtonian liquids. The distribution factors for vertical pipe and annulus are almost identical, which was C 0 = 2.33.
2.3.1 Flow-Pattern Identification for Two-Phase Flow in Annuli

In gas-liquid flow, the two phases can be distributed in the conduit in many configurations called flow patterns. Hydrodynamics of the flow, as well as the mechanisms of momentum, heat and mass transfer, change significantly from one pattern to another. As a result, accurate understanding of any process in two-phase flow depends on the knowledge of the existing flow pattern, and determining the flow pattern is the first step for developing techniques to accurately predict pressure drop, gas and liquid holdup etc. As mentioned previously, only bubble and slug flows occur in annuli during SCP buildup (Casariego, 1981). In our study, we concentrate on these two situations. The experimental data collected by Caetano (1992) reveal that, although the same flow pattern described for pipe occurs in annuli, their characteristics can be different. Fig.2.5 and Fig.2.6 show all the flow patterns in concentric and fully eccentric annuli, respectively.
Bubble Flow. The gas phase is dispersed into small discrete bubbles in a continuous

liquid phase, forming an approximately homogeneous flow through the annulus crosssectional area. The discrete bubbles occur in two different shapes, namely spherical bubbles and cap bubbles. The spherical bubbles are very small, in the order of 3 to 5 mm in diameter, compared to the annulus-cap bubbles, which are relatively larger but still always smaller than half of the configuration hydraulic diameter. The small spherical bubbles follow a zig-zag path, whereas the cap bubbles rise straight and faster. In a fully eccentric

19

annulus, there is a tendency for the small bubbles and cap bubbles to migrate into the widest gap of the annulus cross-sectional area. This causes a higher local void fraction relative to the cross-sectional average void fraction. At high liquid velocities, the mixture appears to flow at the same velocity with no slippage between the phase, regardless of the annulus geometry.

Fig.2. 5 Upward-vertical-flow patterns in concentric annuli (from Caetano, 1992)

Fig.2. 6 Upward-vertical-flow patterns in fully eccentric annuli (from Caetano, 1992)

Slug Flow. The flow is characterized by large gas cap bubbles moving upward,

followed by liquid slugs that bridge the entire cross-sectional area and contain small spherical gas bubbles. The large gas bubbles, which occupy almost the entire cross20

sectional area of the annulus, are similar to the ones occurring in pipe flow and also termed Taylor bubbles. The Taylor bubbles do not occupy the total cross-sectional area (Fig.2.5) because they have a preferential channel through which most of the liquid ahead of the bubbles flow back. This preferential channel exists from the top to the bottom of the bubble and from the inner wall to the outer wall of the annulus. Because of this channel, no symmetry is observed for the Taylor bubble with respect to either vertical or horizontal planes. The liquid phase flows backward as films, around the Taylor bubble, and through the preferential channel, wetting both the inner and outer walls. This tends to create a turbulent region behind the Taylor bubble. For eccentric annuli, the preferential liquid channel is always located where the pipe walls are in contact (Fig.2.6). Comparison among flow patterns occurring in upward vertical flow in a pipe and in an annulus reveals that the existence of an inner pipe in the annulus changes the characteristics of slug flow. The Taylor bubbles in an annulus are not symmetric, having a preferential liquid flow channel through which most of the liquid phase is shed backward. Modification of bubble rising velocity is a function of the pipe diameter ratio and the eccentricity of the annulus. Methods to predict the occurrence of the various flow patterns have been divided into two categories: flow pattern maps and flow-pattern theory.
Flow Pattern Maps. A large number of flow-pattern maps have been generated from

experimental data to correlate the flow pattern in a simple graphical form. Unfortunately, most of the flow-pattern maps have little theoretical basis for the flow-pattern transitions; extrapolation to other fluids, pipe sizes and flow conditions can be dangerous. Nevertheless, flow-pattern maps are used due to their ease of application and also due to

21

the lack of a comprehensive flow-pattern theory until recently. The selection of coordinates for the flow-pattern maps has been of two basic types. Some researchers use dimensional coordinates such as superficial velocity (Sternling 1965, Wallis 1969) or superficial momentum flux (Hewitt and Roberts 1969). Given any single pipe size and set of fluid properties, these coordinates will map the transitions. However, the location of these transition curves may change for changes in these variables. Govier and Aziz (1972) modify these dimensional coordinates for systems other than airwater by considering property ratios between the fluids of interest and that of the air-water system. However, no theoretical basis suggests that the modification correctly generalize the results. Alternatively, the results can be represented using dimensionless coordinates. In the absence of a theoretical basis, the use of dimensionless coordinates is no more general than the use of dimensional ones. Further, one pair of dimensionless groups does not characterize the variety of transitional boundaries that exist. The dimensionless group selected by Duns and Ros (1963) and also used by Orkszewski (1967) and Gould (1974) seem arbitrary. Griffith and Wallis (1961) attempted to invoke theory to arrive at suitable coordinates. They were able to show that the dimensionless coordinate v m gD and vGS vG controlled the transition from the slug to annular patterns. The theory can not provide an analytical expression for the transition curve, and experimental data were used to provide for the unknown constants. As discussed, the use of these same coordinates for the other transitions is open to question. All above discussion is for two-phase flow in pipes. When two phases flow through annuli, flow-pattern can be predicted by use of the hydraulic diameter 22

dh = do - di

(2. 8)

Flow Pattern Theory Dukler and Taitel and their coworkers have published theoretical flow-pattern transitions for horizontal (Taitel and Dukler 1976) and vertical (Dukler and Taitel 1977 and Taitel, Bornea, and Dukler 1980) flow situation. Comparison of their horizontal flow-pattern theory to the Mandhane et al. (1974) flow map is in good agreement. Ishii and his coworkers (Ishii and Mishima 1980, Kataoka and Ishii 1982, and Mishima and Ishii 1983) have also developed theoretical flow-pattern transition for vertical flow. The map of Mishima and Ishii (1983) compares favorably with the Dukler (1977, 1980) maps for vertical flow. Caetano et al (1992) conducted a combined experimental and theoretical study of upward two-phase flow in vertical concentric and fully eccentric annuli. The models proposed to predict flow-pattern transitions in an annulus are similar to those proposed by Ansari et al (1994) for circular pipes (Fig.2.7).

Fig.2. 7 Flow Pattern Map for Pipes (after Ansari, 1994)

23

Bubble/Slug Transition for Flow in Annuli. The minimum equiperiphery diameter for which bubble flow occurs is d EP
s L (r L - r g ) = 19.7 2 rL g
1 2

(2. 9)

where dEP is the equiperiphery diameter defined by d EP = d i + d o (2. 10)

and di and do are the inner and outer diameter, respectively. r L and r g are density of liquid and gas, respectively. s L is surface tension of liquid. For equiperiphery diameter larger than this, transition to slug flow was experimentally found to occur at average gas void fraction equal to 0.20 for flow through a concentric annulus and 0.15 for flow through a fully eccentric annulus. Thus, the Taitel (1980) model was modified for the bubble/slug transition in an annulus by use of values of the gas void fraction measured at this transition. The resulting equation for the transition can then be expressed by v Sg
(r L - r g )gs L v = SL + 0.306 2 4.0 rL
1 4

(2. 11)

for flow through a concentric annulus, and v Sg

(r L - r g )gs L v = SL + 0.230 2 5.67 rL
1 4

(2. 12)

for flow through a fully eccentric annulus. Transition to Dispersed-Bubble Flow. At high superficial liquid velocities, the Barnea (1987) criterion for transition to dispersed-bubble flow was modified with the hydraulic diameter concept. 24

Thus, the transition from either bubble or slug flow to dispersed-bubble flow then becomes
0.4s L 2 (r L - r g )g
1 2

rL s L

f' 2d h

(v

SL

+ v Sg )

1.2

v Sg = 0.725 + 4.15 v SL + v Sg

0.5

(2. 13)

where the hydraulic diameter is defined by Eq.2.8. f is the Fanning friction factor evaluated for the homogenous mixture flowing in either the concentric or eccentric annulus. For a uniform bubble-size distribution and the cubic lattice packing, 0.52 is the maximum gas void fraction allowable under dispersed-bubble conditions. Higher values of void fraction will cause transition to slug flow. Applying this criterion to slip flow yields the transition boundary to dispersed-bubble flow at gas void fraction equal to 0.52. v Sg = 1.083v SL
(r L - r g )gs L + 0.796 2 rL
1 4

(2. 14)

2.3.2 Gas Slip Velocity

The rise velocity of isolated bubbles in stationary column of liquid, i.e., the gas slip velocity, is an important parameter for the estimation of the gas in-situ velocity when bubbling gas flows through stagnant liquid. It also depends on the flow-pattern.
Bubble Flow. Harmathy (1960) developed a correlation for experimental data to

describe the rise of single, slightly large, bubbles as a function of density difference and surface tension. This correlation, which is independent of bubble size, gives
g ( r L - r G )s v s = 1.53 2 rL
1 4

(2. 15)

25

To account for the effect of bubble swarm, Zuber and Hench (1962) modified the expression by liquid holdup H L :
g ( r L - r G )s v s = 1.53 2 rL
1 4 n' HL

(2. 16)

where the value of n varies from one study to another. Ansari (1994) found that a value of 0.5 for n gave the best results. Wacholder (1973) employed a statistical averaging technique (Barchelor, 1972) to estimate the terminal velocity of bubbles or drops when their volume fraction c is small. His analysis has been corrected recently by Haber and Hetsroni (1981); who gave the following expression for the terminal velocity of gas bubbles:
Uc = 1 - 4.49c + o(c 2 ) U

(2. 17)

where Uc is a velocity of any representative drop in the swarm and U is the corresponding Hadamard-Rybczynski (1911) velocity for an isolated drop: U= 2rB 1 + k (r l - r g )g 3m 2 + 3k
2

(2. 18)

m is the ratio of viscosities inside and outside the drop. rB is the bubble where, k = m diameter. When k = 0, Eq.2.18 is the velocity of a gas bubble in extended fluid. Determination of the coefficient of o(c 2 ) in Eq.2.17 using the statistical technique is extremely difficult and has not met with much success to date. Hence these technique have been limited to dilute dispersions (c << 1). Sangani and Acrivos (1983) have considered the case of a dispersion in which particle centers coincide with the lattice points of a periodic array. Sangani and Acrivos (1983) employed a method of multipole expansions and obtained the terminal velocities of freely 26

circulating gas bubbles (k = 0). Their results are presented in Table 2.2 in terms of a hindrance factor YH defined by r U c = B (r l - r g )g YH 3m where, r l = liquid density
r g = gas density
2

(2. 19)

rB = radius of gas bubble

m = liquid viscosity

Table 2. 2 - The Hinderance Factor YH (cf. Eq.2.19)

c YH Freely circulating SC 1.140 1.233 1.396 1.609 1.898 2.315 2.967 4.110 5.060 6.510 8.840 12.80 BCC 1.117 1.266 1.46 1.724 2.101 2.678 3.65 5.54 7.29 10.26 15.99 28.7 FCC 1.121 1.276 1.48 1.76 2.168 2.808 3.927 6.26 8.59 13.01 23.34 59 SC 1.747 2.082 2.25 3.228 4.263 5.961 9.006 15.09 20.46 28.74 41.85 63.1 YH Noncirculating BCC 1.779 2.173 2.75 3.635 5.83 7.662 13.297 25.35 38.7 64.2 118 243 FCC 1.789 2.2 2.811 3.765 5.362 8.306 14.575 31.3 51.9 97.6 228 657

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.85 0.9 0.95 1

Note: SC - Simple Cubic Array, BCC - Body-Centered Cubic Array, FCC - Free-Centered Cubic Array.
c in that table is defined as
c c = c max
13

(2. 20)

where cmax is the maximum volume fraction, i. e. the volume fraction of bubbles when they are touching each other, whose value equals

p 6 = 0.5236 ,

3p 8 = 0.6802 ,

2p 6 = 0.7405 , respectively for a simple, body-centered, and face-centered cubic array.

27

Their results are plotted in Fig.2.8 where the terminal velocity is relatively insensitive to the geometry of the array provided that c < 0.3.

70

60

50

SC BCC FCC

40

YH
30 20 10 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

Fig.2. 8 YH vs. volume fraction for slightly large bubbles Sangani and Acrivos also gave asymptotic results for small volume fractions:
1 - 1.1734c1 3 - 0.1178c 2 + o c 8 3 = 1 - 1.1946c1 3 + 0.3508c 2 + o c 8 3 1 - 1.1945c 1 3 + 0.3611c 2 + o c 8 3

-1 YH

( ) ( ) ( )

(BCC ) (FCC )

(SC )

(2. 21)

The results agree with the numerical results presented in Table 2.2 within 5% for c < 0.6. The preceding results for freely circulating bubbles are expected to apply for relatively large bubbles (Eo >> 1). Small bubbles (Eo << 1) would tend to behave like rigid particles. The Eotvos number Eo is
Eo = rB2 (r L - r g )g
g

and g is interfacial tension. Sangani and Acrivos (1982) have also computed the terminal velocities of rigid particles and their results are presented in Table 2.2 along with those of freely circulating 28

particles (Fig.2.8 and Fig.2.9). From these figures we note that the hindrance to rise velocity of a bubble in the presence of other bubbles is much more pronounced in case of rigid interfaces. We also can obtain expressions for noncirculating bubbles ( k = ) from
Table 2.2.
3.3057 - 24.952c + 255.93c 2 YH = 2.45531e ( 6.7276 c ) 2.2601e ( 7.3013c )

( SC ) ( BCC ) ( FCC )

(2. 22)

700 600 500 400 SC BCC FCC

YH
300 200 100 0 0 0.1 0.2 0.3 0.4

0.5

0.6

0.7

0.8

Fig.2. 9 Asymptotic results of YH vs. volume fraction for small bubbles Marrucci (1965) proposed an expression relating the velocity of rise of a swarm of shperical bubbles to the velocity of a single bubble. It is given by:
v sw = v
2 HL

1-a

1 < N Rb < 300

(2. 23)

where, v sw = velocity of a swarm of bubbles with respect to the liquid v = velocity of a single bubble in an extended liquid

a = volume fraction of gas

HL = 1-a

29

Bhatia (1969) derived a relationship between volumetric gas fraction and bubble velocity:
v sw = v tanh 0.25(1 a )

(2. 24)

Bhavaraju, Mashelkar and Blanch (1970) made a theoretical study on the motion of a swarm of bubbles in a power law fluid. They concluded that the bubble swarm velocity increase as the pseudoplasticity of the fluid increases for a given holdup. This is opposite to the trend for single bubble velocity. Using minimum energy dissipation principle, Carsariego (1987) developed a theoretical relationship between gas concentration and the average upward swarm velocity for migration of a gas-contaminated region. Hasan and Kabir (1992) concluded that the terminal gas rise velocity in annuli, vs, was not affected significantly by either the inner pipe diameter or the pipe-deviation angle from the vertical. Therefore, they suggested that Harmathys equation (Eq.2.15) be used in bubble flow in annuli. Considering the effect of bubble swarm, Caetano et al (1992) suggested Ansaris equation (Eq.2.16) with an iterative process to calculate the liquid holdup H L and gas slip velocity vs.
Slug Flow. Assuming that liquid slugs do not contain any gas bubbles, the gas slip

velocity is equal to the rise velocity of a Taylor bubble in a stagnant fluid. For large bubbles that almost fill the pipe, Davies and Taylor (1950) considered inviscid flow around the bubble nose. They derived the equation v s = 0.35 g ( r L - r G )d r L (2. 25)

where d = pipe diameter. This is normally referred to as the Taylor bubble velocity. 30

Hasan and Kabir (1988) derived the rise velocity of a Taylor bubble in an inclined pipe as v s = 0.35 g (r L - r G )d 1.2 sin q (1 + cosq ) rL (2. 26)

There are a limited number of studies on annuli. Caetano (1995) investigated experimentally and analytically the rise velocity of a large single bubble in Newtonian fluids inside annuli. He calculated the rise velocity of a Taylor bubble by using the equiperipheral-diameter concept:
v s = 0.345 g (d i + d o )(r L - r g )

rL

(2. 27)

Also, using air and water, Hasan and Kabir (1992) investigated two-phase flow in vertical and inclined annuli. They stated that their Taylor-bubble-rise velocity showed a linear relationship with the diameter ratio, d i d o , suggesting the following expression for the Taylor-bubble-rise velocity for inclined annuli

d i g (r L - r G )d o 1.2 vs = 0 . 345 0 . 1 sin q (1 + cosq ) + do rL

(2. 28)

Studies on bubble migration in non-Newtonian fluids inside annuli are rare. Rader et al. (1975) proposed a correlation for migration rate considering viscous effects, annulus configuration, gas bubble expansion and liquid velocity effects. However, they did not include in their correlation experimental results obtained with non-Newtonian fluids. Based on their data and the results of Radel et al. (1975), Koederitz (1976) developed a gas rise velocity correlation in a static fluid column. It is given by:
v s = [0.16 + 0.092 log10 ( N Re B )]

(d i + d o )( r L - r g ) rL

(2. 29)

31

where v s is in feet per second, d i , d o are in inch, r L , r g are in pounds per gallon. Bubble Reynolds number for non-Newtonian fluids is given by

N Re B

133632 r L v s( 2- n ) 0.0208(d o - d i ) n [ ] = 2 + 1n K

(2. 30)

where K is the consistency index in equivalent centipoise and n is flow-behavior index. Santos et al (1997) proposed a semi-empirical correlation for estimating large bubbles (Taylor bubbles) rise velocity in non-Newtonian fluids. The correlation accounts for the non-Newtonian behavior of the drilling fluid systems, annulus inclination and geometry of the annular space. It is expressed as
v s = C1C 2 C 3 gd o (r L - r g ) r L

(2. 31)

Constant C1 accounts for the effects of annulus geometry. It is obtained by linear fitting the experimental data from their paper and others.

C 1 = 0 . 3143

di + 0 . 2551 do

(2. 32)

The constant C2 accounts for the non-Newtonian behavior of the drilling fluid. It has been correlated to bubble Reynolds number. The generalized bubbles Reynolds number RNb is defined as:
N Re B

9.782v s2- n r L = K

do - di ( 4 2 + 1 n )

(2. 33)

The constant C2 is:

C 2 = 0 . 0532 log

10

(N

Re

)+ 0 . 7708

(2. 34)

The constant C3 accounts for the inclination effect. It is expressed as: C 3 = 1.0 + F sin(2a )
32

(2. 35)

where F had been correlated to the generalized bubble Reynolds number:

F = 0.0586 log10 ( N Re B ) + 0.0042

(2. 36)

Both Keoderitz and Santos correlation need an iterative procedure to obtain the correct value for the gas rise velocity, because it is included in the calculation of bubble Reynolds number.
2.3.3 Gas Distribution in Mud Column above Cement

During a shut-in test, Wellbore Phase Redistribution (WPR) frequently occurs in wellbore having very compressible fluids, such as low-pressure, single-phase gas and high gas-oil ratio mixture, including steam. Phase redistribution is the relative movement of gas phase in wells production string after shut-in at the surface. The preferential movement of the gas phase is primarily due to the buoyancy and channeling tendency of the gas bubbles. Stegemeier and Matthews (1958) reported phase redistribution and its consequent impact on pressure buildup test. The pressure buildup curve rose to a maximum and then declined to the stabilized reservoir pressure. They attributed the pressure rise to phase redistribution or segregation. Pitzer et al (1959) also reported observing the characteristic hump on the buildup curve of a well with surface shut-in. Stegemeier and Matthews (1958) observed the predominance of the phase redistribution phenomenon in wells with large positive skin and in reservoir with moderate permeability. Fair (1979) concluded that WPR was a wellbore storage effect. Assuming the exponential pressure buildup, he applied methods for wellbore storage to analyze and describe the anomalous pressure buildup. We found that the simulation of WPR has several similarities with that of SCP. In both situations, there is a multiphase flow in mud after the closure at top. Also surface and bottom pressures are affected by the combined effect of afterflow and bubble migration.

33

However, the most obvious difference between the two is the fluid that flows in porous media. In WPR, single-phase oil flow or gas-liquid multiphase flow is assumed. While in SCP, single-phase gas flow is assumed in cement. Nevertheless, the techniques used in simulating WPR inspired a numerical model to describe pressure and gas distribution throughout mud column in SCP diagnostic tests.
Hybrid Approach. Several hybrid approaches were used to couple the wellbore

(numerical) with the reservoir (analytical) in WPR. Assuming single-phase oil flow in reservoir, the sandface flow rate can be obtained by Duhhamels principle or Meunier et als rate-convolution time function. Hasan and Kabir (1992) developed a simulator for WPR that accounts for continually decreasing afterflow from the reservoir since shut-in, the variation of void fraction within the wellbore, and the combined effect of afterflow and bubble migration on the wellhead and bottomhole pressure. Assuming that only oil flows in the reservoir, they calculated sandface flow rate by adapting the work of Meunier et al (1985). qD = q sf q0
=

1 pi - p ws - log(t + Dt ) - (Dt ) s m

(2. 37)

where the skin coefficient, s, semilog slope, m, and the sandface rate-convolution time function, (Dt), have been defined by Meunier et al (1985). To account for fluid movement within the wellbore, they divided the wellbore into N number of cells as shown in Fig.2.10. They also assumed that a gas chamber is instantaneously formed when the well is shut-in at surface. This top cell is completely filled with gas (no liquid) and it cannot lose gas to any other cells although it receives gas from the cell immediately below it. 34

Gas 1

i-1

Out
i

Vg(tj) = Vcfg(tj) In
i+1

n-1 n

Fig.2. 10 Schmetic representation of the wellbore for cell material balance Therefore, the wellhead pressure at any time step, tj, is related to the pressure at previous time step, tj-1, through the gas law.
p wh (t j ) = p wh (t j -1 )n(t j )Vwh (t j -1 )z (t j ) n(t j -1 )Vwh (t j )z (t j -1 )

Vg(tj) = Vg(tj-1) + (C0vm+v)(Dt)[Vgi+1 (tj-1) ] - (C0vm+v)(Dt)[Vgi (tj-1) ]

(2. 38)

Using an iterative procedure, they calculated wellhead and bottomhole pressure buildup after the surface shut-in. They validated the simulator with limited simulations. Hasan and Kabir (1993) developed a simplified mathematical model from the motion of a single bubble (spherical or Taylor). The model helps to understand principal causes for WPR. Their results show that both the wellhead and bottomhole pressure increase with time as a single gas bubble ascends up the liquid column. They concluded that the magnitude of the anomalous pressure rise increases with increasing skin damage, decreasing wellhead pressure, and increasing bubble rise velocity. 35

Xiao et al developed a mechanistic model to simulate WPR. Their model rigorously treats wellbore and reservoir flow interaction, and handles the effect of inter-phase mass transfer. Assuming single-phase radial flow of oil in the reservoir, they obtained the boundary condition at sandface by applying Duhamels principle, i.e.,
p wsD = kh( pi - p ws ) = 141.2qm
t pD + Dt D

q sf (t ) q

' (t pD + Dt D - t )dt + s p cD

q sf (t pD + Dt D ) q

(t

pD

+ Dt D ) (2. 39)

where q is the well sandface flow rate before shut-in. tpD is the dimensionless producing time and pcD is the dimensionless pressure solution at the sandface for the constant sandface rate problem with no skin effect, i.e., s = 0, which is pcD = kh( pi - p wf 141.2qm

1 (ln t D + 0.80907) 2

(2. 40)

They consider gas-liquid two-phase flow in the wellbore. And the gas is released after the in-situ pressure is lower than the bubble-point pressure. The black-oil formulation is applied to account for the interphase mass transfer. Since phase segregation after the shutin is a slow and gravity dominated process, they assume that the flow of the gas-liquid mixture is in local equilibrium and the fluid-wall friction is negligible. A simplified mixture momentum equation is used. r g p =- m z 144 g c (2. 41)

Since the effect of phase segregation is maximized for bubble and slug flow (Pitzer et al, 1959), they only considered those two flow pattern to calculate gas slip velocity. They assumed that bubble exists for Eg 0.15, slug flow exists for Eg 0.25. In the between, the gas slip velocity is obtained by linear interpolation of bubble rising velocity and slug 36

rising velocity. The whole procedure is also iterative. They used oil component concentration in each cell as the criteria for convergence. They also developed a numerical scheme to implement and verify the model.
Numerical Modeling Approach. Several numerical models have been developed to

study the effect of multiphase flow in the wellbore and in the surrounding formation during well testing. Winterfeld (1989) presented a simulator with multiphase flow in the wellbore to study the phase redistribution phenomenon. He discretized the wellbore and the reservoir in gridblocks and used the Hagedorn and Brown (1965) two-phase flow correlation. Winterfeld (1989) reported satisfactory simulation of an actual well test affected by wellbore phase redistribution. Almehaideb et al (1989) applied the semi-empirical approach presented by Winterfeld (1989) to handle phase to phase and phase to wall viscous terms. Their results showed that the two-fluid model of wellbore fluid can simulate a pressure hump if the liquid in the well flows back into the reservoir. But the mixture model of two-phase flow is unable to generate a pressure hump in the same situation. If no appreciable backflow occurs, both models give basically the same results. Numerical modeling approach can solve complicated problem such as two-phase flow in porous media. But the calculation time will be long. A Hybrid approach couples the existing analytical solution with numerical solution. In this study, a hybrid method is used. But pressure distribution in was the analytical solution of a linear gas flow in cement and numerical solutions of two-phase flow in annuli described the gas migration above the cement.

37

2.3.4 Frictional Pressure Loss

Without a significant liquid flow, friction pressure loss in SCP is mainly caused by bubble swarms moving against annular walls. Many different correlations for friction factor have been developed over the years. Usually, homogeneous flow conditions (no slip or no velocity difference between the phase) are assumed to derive it. The friction-pressure-gradient for annuli is given by 2 f ' r Tp v m dp = do - di dL f
2

(2. 42)

where the mixture velocity, vm, is the sum of superficial velocity of liquid v SL and gas v Sg : vm = qL qg + = v SL + v Sg A A (2. 43)

In the Eq.2.42, only mixture density r TP used slip liquid holdup H L :

r TP = r L H L + r g (1 - H L )

(2. 44)

The Fanning friction factor f, is determined by methods in following sections using Reynolds number for two-phase flow N ReTP = r TP v m d h m TP (2. 45)

To use those existed equations for Newtonian flow, apparent Newtonian liquid viscosity is used (Metzner and Reed, 1955) for two-phase flow in power-law fluid.
4n 8v m ma = K 3n + 1 d o - d i
n -1

(2. 46)

The mixture viscosity for two-phase flow in non-Newtonian fluid, m TP , is

m TP = m a l L + m g (1 - l L )

(2. 47)

38

where the no-slip liquid hold-up, lL , is given by

lL =

v SL v SL + v Sg

(2. 48)

Two geometrical parameters identify the configuration of an annulus: the annulus pipe-diameter ratio, K ' = d i d o , and the degree of eccentricity. The degree of eccentricity accounts for the displacement of the inner pipe center from the outer pipe center: e= 2 DBC (d o - d i ) (2. 49)

where DBC = distance between the pipe centers. Eccentricity changes from zero to one.
Laminar Flow. A rigorous treatment of the flow field is possible for any annulus

configuration for laminar flow, where Reynolds number is less than 3000. Bird et al. (1960) gave analytical solutions for both the velocity profile and friction factor for a concentric annulus. The Fanning friction factor for a concentric annulus is
f
' CA 2 FCA ( 16 1 - K ') = = N ReTP N ReTP 1 - K ' 4 1 - K ' 2 2 ln (1 K ') 1 - K '

(2. 50)

Thus, FCA = FCA (K ) = 16(1 - K ')

2

1 - K ' 4 1 - K ' 2 2 ln (1 K ') K 1 '

(2. 51)

where FCA is the friction-geometry parameter for a concentric annulus. By using a bipolar coordinate system, Snyder and Goldstein (1965) presented an analytical solution for laminar flow in an eccentric annulus. The solution was based on previous development by Heyda (1959) and El-Saden (1961). Applying the Fanning equation, it can be shown that

39

' f EA =

2 FEA 1 4(1 - K ') 1 - K ' 2 = N ReTP N ReTP f sinh 4 h 0

(2. 52)

and
FEA = FEA (K ' , e ) =

4(1 - K ') 1 - K ' 2 f sinh 4 h 0

2

(2. 53)

where
1 2n 2 f = (cosh h i - cosh h o ) 2 2 nh i h -h - e 2 nho n =1 e o i

1 1 1 + 4 sinh 4 h - sinh 4 h o i

(2. 54)

cosh h i =

K ' 1 + e2 - 1 - e2 2K ' e

) (

) )

(2. 55)

K ' 1 - e2 - 1 + e2 cosh h o = 2K ' e

) (

(2. 56)

Turbulent Flow. Combining the Cunn and Darling (1963) developments into a

Nikuradse-type expression, the friction factor for concentric and eccentric annuli are predicted, respectively, from 1
(N Re TP -3, 000 ) 0.45 exp 10 6 ' FP f CA F CA

(2. 57)
1 2

-3, 000 ) (N 0.45 exp - ReTP 10 6 ' FP = 4 log N ReTP f CA F CA for concentric annuli, and

- 0.4

40

1
-3, 000 ) (N 0.45 exp - ReTP 10 6 ' FP f EA F EA 1 2

(2. 58)
1 2

- 3, 000 ) (N 0.45 exp - ReTP 106 ' FP = 4 log N ReTP f EA F EA

- 0 .4

for eccentric annuli. In these equation, f = Fanning friction factor, FCA, EA = laminar flow friction geometry parameter for concentric and eccentric annuli, and Fp = 16 for pipe flow. 2.4 Flow Through Chokes According to present MMSs policy, only casing pressure history and amount of removal liquid need to be recorded in bleed-down. Using pressure differences between upstream and downstream of the valve, gas and liquid flow rates can be calculated, which are important upper boundary conditions in modeling pressure bleed-down. Fig.2.11 shows a typical needle valve. Usually, it has four orifice sizes available, such as 1/32, 1/16, 3/32, 1/8.

Fig.2. 11 Needle valve 41

The flow of fluids through the needle valve can be considered as flow through a choke. The schematic of a choke is shown in Fig.2.12.

P1

dch

d1

P2

Fig.2. 12 Choke schematic For compressible flow, it is possible for the velocity in the choke throat to reach the speed of sound or the sonic velocity in the fluids. If the fluids reach sonic velocity within the choke, the flow behavior becomes independent of conditions downstream from the choke. This condition is called critical flow. If the maximum velocity of fluids in the choke is less than the sonic velocity, the flow is called subcritical flow. Fig.2.13 shows the dependence of flow rate through a choke on the ratio of the downstream to upstream pressure for a compressible fluid. As the pressure ratio decreases, flow rate increases. When the pressure ratio reaches a critical value, the flow rate through the choke becomes constant.
500 400 Gas Rates, Mscf/D 300 Subcritical Flow 200 100 0 0 0.2 0.4 0.6 0.8 1 1.2 Pressure Ratio, y Critical Flow 1/8 in.

Fig.2. 13 Dependence of choke flow rate on y (after Beggs, 1984)

42

During the bleed-down, single-phase gas or multiphase gas and mud flows through the 0.5 needle valve. Therefore, this study concentrates on these two flows. 2.4.1 Single-Phase Gas Flow For gases, Bernoullis equation can be combined with an isentropic (adiabaticfrictionless) equation of state. All irreversible losses are accounted for through a discharge coefficient. The resulting Eq.2.59 (Beggs, 1984) is applicable for both critical and subcritical flow. However, for critical flow, the pressure ratio y = p 2 p1 is replaced by the critical-pressure ratio, yc.

q sc = where

2 C n p1 d ch

g g T1 Z 1

2 k +1 k k y - y k k - 1

(2. 59)

Cn =

C s C D Tsc p sc

(2. 60)

Values of the constants in Eq.2.59 and Eq.2.60 for both customary and SI units was given in Table 2.3. Table 2. 3 Constants and Units for Eq.2.59 and Eq.2.60 Symbol Customary SI Metric qsc Mscf/D m3/d dch in. mm p psia kPa K T R 27.611 1.6259 Cs 0.865 0.865 CD 14.696 psia 101.325 kPa psc 273.16 K Tsc 591.68 R 844.57 3.7915 Cn The critical-pressure ratio for a gas with a ratio of specific heats k = Cp/Cv is give by p2 2 k -1 yc = = p 1 c k + 1
43
k

(2. 61)

For air and other diatomic gases, k is approximately 1.4 and the critical-pressure ratio from Eq.2.61 is 0.53. For practical reasons, critical flow for gases is often estimated to occur at a pressure ratio of 0.5, i.e., the upstream pressure equals two times the downstream pressure. Some manufacturers (such as Ideal Aerosmith Inc.) found that the calculated flow is usually much higher than measured value. Instead, the following equations are used to calculate the gas rate through the needle valve.

q sc = 816CV

P1
g g T1

(2. 62)

for critical flow (p1 2p2); and

q sc = 962CV

(p

2 1

2 - p2 g g T1

(2. 63)

for subcritical flow ( p1 2 p 2 ).

where
qsc = Gas Flow in SCF/H T1 = Absolute temperature in R. (F + 460) p1 = Upstream (inlet) pressure in psia p2 = Downstream (outlet) pressure in psia
gg = Specific Gravity of medium where air at 70 F and 14.7 psia = 1.0

The coefficient of flow (Cv) is a formula used to determine a valves flows under various conditions and is thus useful for selecting the correct valve for a flow application. The Cv was designed for use with liquid flow. It expresses the flow in gallons per minute of 60 F water with a pressure drop across the valve of 1 psi. However, this same Cv value

44

can be used to determine gas flows through a valve. Usually manufacturers will create flow charts which display the Cv value at different turns for calculating flows for varying conditions at any of these points (Table 2.4 and Fig.2.14). Table 2. 4 Flow Characteristics of Orifice Sizes (after Ideal Aerosmith Inc.) -1- (.031); -2- (.062); -3- (.094); -4- (.125) No. Turns Cv Open -1-2-3-41 2 3 4 5 6 7 8 9 10 12 14 16 18 20
0.25 1/32" 1/16" 3/32" 1/8"

0.0007 0.0017 0.0027 0.0037 0.0047 0.0057 0.0067 0.0077 0.0087 0.0107 0.0127 0.0148 0.0169 0.0190

0.0053 0.0120 0.0184 0.0245 0.0303 0.0358 0.0410 0.0458 0.0504 0.0546 0.0623 0.0688 0.0742 0.0786 0.0818

0.0176 0.0353 0.0509 0.0644 0.0762 0.0863 0.0948 0.1021 0.1081 0.1130 0.1204 0.1254 0.1292 0.1331 0.1383

0.0236 0.0443 0.0635 0.0814 0.0980 0.1133 0.1273 0.1402 0.1520 0.1628 0.1812 0.1960 0.2075 0.2162 0.2224

0.2

0.15 Cv 0.1 0.05 0 0 5 10 No. of Turns Open 15 20

Fig.2. 14 Flow chart of needle valve (after Ideal Aerosmith Inc.) 45

2.4.2 Multiphase Flow

Determination of the boundary between critical and subcritical flow is very important to predict the flow behavior for multiphase flow through chokes. The proper choice depends on whether a calculation is made of the critical-pressure ratio below which totalmass flow rate is constant, or if the sonic velocity of a mulitphase mixture is estimated.
2.4.2.1 Critical-Flow Boundary

Several authors have developed methods to predict the critical-flow boundary in multiphase flow. Ashford and Pierce (1975) developed an expression for total-mass flow rate of a multiphase mixture. They assumed isentropic flow through the restriction, an incompressible liquid, no liquid flashing in the choke and a homogenous mixture. Eq.2.64 assumes the derivative of the flow rate with respect to pressure ratio is zero at the critical boundary.
-e R1 b 1 - y c - y c + 1 y c - 1 = 0 1 - b k + k 1 R y 1 c 2 R1
2

(2. 64)

where, b =

k -1 k +1 and e = . k k

Eq.2.64 requires an iterative procedure to determine values of yc as a function of in-

situ gas/liquid ratio and k. In-situ gas/liquid ratio at upstream conditions, R1, can be calculated easily as the ratio of the superficial gas to liquid velocities determined at conditions immediately upstream of the choke. Fig.2.15 gives a plot of Eq.2.64 and shows that at values of R1 above 10, the critical-flow boundary is similar to single-phase gas. However, for lower valves of R1, significantly lower values of yc are predicted. 46

0.7 0.6 0.5 0.4

yc, critical pressure ratio, p1/p2

Subcritical Flow

Critical Flow
0.3 0.2 0.1 0 0.001
k = 1.04 k = 1.20 k = 1.40

0.01

0.1

10

100
3 3

1000

10000

100000

R, gas/liquid ratio, ft /ft

Fig.2. 15 Critical-flow boundary correlation for multiphase chokes (after Ashford, 1975) Sachdeva et al.(1986) performed a combined experimental and theoretical study and resulted in a dimensionless equation to determine yc for oil, gas and water flowing through a choke. Determining yc from the equation requires an iterative procedure. Perkins (1993) derived a complicated equation to predict the critical-pressure ratio that follows very closely the approach of Ashford and Pierce (1975). He combined equations for conservation of mass and isentropic expansion for total mass flow rate using the same assumptions as Ashford and Pierce in Eq.2.64. Also, an iterative procedure is required to determine yc. Fortunati (1972) presented an empirical method that can be used to calculate both critical and subcritical multiphase flow through chokes. He assumed a homogeneous mixture and contended that the assumption was valid as long as v m is greater than 32.8 ft/sec and the mixture Froude number is greater than 600. 47

Wallis (1969) presented an expression to calculate the sonic or compressibility-wave velocity of a homogeneous mixture. Wallis concluded that the sonic velocity of a homogeneous mixture passes through a minimum at a no-slip void fraction of 0.5. Nguyen et al. (1981) studied sonic velocity in two-phase systems as a function of flow pattern. He gave a sonic velocity for idealized slug flow and homogeneous flow respectively. But for stratified flow, he proposed that an effective sonic velocity existed in each phase that is influenced by the other phase.
2.4.2.2 Subcritical-Flow Behavior

Experimental and field tests confirm that accurate prediction of subcritical flow rate through restrictions is very difficult. Following extensive tests of subcritical two-phase flow through velocity-controlled subsurface safety valves, a simple homogeneous Bernoulli-type equation was adopted. It is called TUFFP model.
Dp =
2 r n1v mB 1 = p1 - p 2 (atmospheric pressure ) 2 2g cCD

(2. 65)

where,
v mB1 = qL + qg p 2 d ch 4

r n1 = r L1l L + r g1 (1 - l L )

p1 and p 2 respectively are upstream and downstream pressure in psia. q L and q g

respectively are liquid and gas flow rate in ft3/sec. d ch is choke size in 64ths of an inch. r n1 is the mixture density in lbm/ft3. r L1 and r g1 respectively are liquid and gas density calculated at upstream pressure (p1). In this model, CD = 0.5 and gc = 32.2. 48

Sachdeva et al. (1986) also presented a more complicated equation to calculate flow rate through a choke.
2.4.2.3 Critical-Flow Behavior

Critical flow exists if y < yc or if mixture velocity is equal to the sonic velocity. Omana et al (1969) presented an empirical correlation to predict critical multiphase flow through Thornhill-Craver-type chokes (Fig.2.12). In his equations, the units should be: qLsc = STB/D, sL = dynes/cm, and p1 = psia.
Nr = rg rL

(2. 66)
1 p1 r s L L
0.5

N p1 = 1.74 10

-2

(2. 67)

N D = 0.1574d ch

rL sL

1.25

(2. 68)

N qL

rL = 1.84q Lsc s L

(2. 69)

Through the application of a least-square regression-analysis procedure to a series of high-pressure natural gas/water tests, the empirical correlation is
.19 0.657 1.80 -3.49 N qL = 0.263N r N3 ND p1 l L

(2. 70)

Nind (1964) stated that a generalized expression for the flow of gas and oil through a knife-edged choke is given by p1 =
.5 Cq Lsc R 0 p 2 d ch

(2. 71)

where, Rp is gas/liquid flow rate ratio Rp = qg q Lsc (2. 72)

49

The expression has been the basis for several modifications by use of experimental and field data. The generalized equation is p1 = bq Lsc R c p
a d ch

(2. 73)

and gas flow rate can be obtained from above equation as qg =

c p1 d ch -c bq 1 Lsc

(2.73a)

where p1 is in psia, qLsc is in STBL/D, Rp is in scf/STBL, and dch is in inches. Table 2.5 gives values of a, b and c proposed by different investigators. Table 2. 5 Choke Constants Investigator a b Ros (1960) 2.00 4.25 10 -3 Gilbert (1954) 1.89 3.86 10 -3 Baxendell (1967) 1.93 3.12 10 -3 Achong (1961) 1.88 1.54 10 -3 c 0.500 0.546 0.546 0.650

Using Sachdeva (1986) model, critical flow occurs when y < yc. The flow rate should be calculated with y = yc and that all fluid properties should be evaluated with p 2 = y c p1 .

50

CHAPTER 3. CASING PRESSURE BUILDUP IN WELLS WITH GAS-FREE LIQUID COLUMN ABOVE CEMENT
3.1 Model Description A model for casing pressure buildup in the annulus with a liquid column above the cement was developed, assuming that the transient period is composed of numbers of short time steps. In each time step, gas flow is assumed to be steady-state in the cement and reaches the top immediately after it releases from the cement, no entrapment of gas in liquid column. In fact, when gas flows in low viscosity, Newtonian liquid, its rising velocity is high and almost no gas will be detained in liquid. The formation pressure is assumed to be constant due to the high formation permeability compared with that of cement. Gas gravity is neglected. The mud is slightly compressible and has constant density. The annulus is incompressible. Temperatures at cement top and casing head ( Twb and Twh ) are considered respectively. Fig. 3.1 shows the conceptual gas migration in annulus.

Fig.3. 1 Gas migration in annuli with a gas-free liquid column above cement 51

Based on these assumptions, the pressure on the wellhead p t in time step n is:
n k 2 4 T wh p sc q c Dt n -1 n -1 n -1 Vt Vt 1 n -1 n k =1 + pt = p t pt + n -1 n -1 n -1 2 c mVm c mVm c mVm Twb

(3. 1)

The pressure on cement top pc can be related to the surface casing pressure by: pcn = ptn -1 + 0.052 r m Lnf -1 The steady-state flow rate on the top of cement is expressed as: (3. 2)

q cn =

0.003164kTsc A 2 p f - p cn p scTLc m i Z i

( )]
2

(3. 3)

The derivation and procedure of calculation is in Appendix B.

3.2 Parameters Affecting Pressure Buildup

Using the model, critical parameters controlling pressure buildup were analyzed. Among those parameters, some are ready to record, such as mud compressibility and formation pressure if the depth of gas source formation can be determined. While others are difficult to monitor, such as the size of gas chamber, annular conductivity.
Size of Gas Chamber Effect. In this study, the gas chamber represents the void

between liquid column top and the well head. Usually, it is filled with gas or gas-cut mud with high gas concentration. Using the model, theoretical pressure buildup curves for different gas chamber volumes was plotted. Because of the compressibility difference between gas and mud, the gas chamber functions as a stabilizer. The larger the chamber, the slower the casing pressure will reach to the stable pressure (Fig.3.2). Moreover, the stabilizing effect is more obvious in the early stage than in the late one. Therefore, keeping a gas chamber after bleed-down may elongate the pressure buildup
52

cycle. On the other hand, by matching the pressure history, the size of the gas chamber can be obtained and the fluid level can be determined readily.
2500

Casing Pressure (psia)

2000

1500

1000
Vc = 0 cu ft Vc = 40 cu ft Vc = 200 cu ft

500

0 0 100 200 300 Time (day) 400 500

Fig.3. 2 Effect of the size of gas chamber (cm = 1.5e-61/psi, kA = 2 md - sq ft, pf = 3500 psia)
Mud Compressibility Effect. It can be observed that the more compressible the

mud, the slower the casing pressure builds up (Fig. 3.3,). Although mud compressibility is easy to measure, it is rarely collected due to a clear use has not yet developed.
2500 2000 1500 1000 500 0 0 50 100 150 200 Time (day) 250 300 350
cm = 1.5E-6 1/psi cm = 1.5E-5 1/psi cm = 1.5E-4 1/psi

Casing Pressure (psia)

Fig.3. 3 Effect of mud compressibility (Vg = 0 cu ft, kA = 3 md - sq ft, pf = 3500 psia) 53

Effect of Annular Conductivity. Quality of cement bond is represented by annular

conductivity, which is the product of cement permeability and annular cross-area. Fig.
3.4 shows that early buildup is very sensitive to this parameter. A small increase in

conductivity leads to a large jump in pressure. The annular conductivity is a hard-to-get data in practice. Using the model, the parameter can be obtained and thus an understanding of cement quality.

2500

Casing Pressure (psia)

2000

1500

1000
KA = 0.1 md - sq ft

500

KA = 0.01 md - sq ft KA = 0.003 md - sq ft

0
0 10 20 30 40 50 60 70 80 90

Time (month)

Fig.3. 4 Effect of annular conductivity (Vg = 0 ft3, cm = 1.5e-6 1/psi, pf = 6500 psia)

Effect of Formation Pressure. Compared with the SCP, the formation pressure is

relatively constant because formation permeability is much larger than the cement permeability. Its magnitude will affect the stable SCP at which no gas migrates from the formation due to the pressure balance. From Fig. 3.5, it can be observed that the higher formation pressure, the higher the stable pressure. 54

3000

Casing Pressure (psia)

2000

1000
Pf = 7000 psia Pf = 6615 psia Pf = 6000 psia

0 0 10 20 Time (month) 30 40

Fig.3. 5 Effect of formation pressure (Vg = 7 ft3, cm = 4e-6 1/psi, kA = 0.3 md-ft2)
3.3 Model Validation with Data from Wells

Field data from two wells were chosen to validate the model. The well geometry and fluid properties are listed in Table 3.1. All those parameters are inputs for calculating theoretical pressure buildup using the model. Table 3. 1 Well Geometry and Fluid Propoties Well 23 Well 24 Twb R 575 552 T R 630 584 R 520 520 Twh D1 ft 0.829 0.829 D2 ft 0.583 0.635 ft 1821 3217 Lc Initial Lm ft 8273 6433 Initial Lg ft 27 0 cp 0.02 0.015 mg psi-1 4.0e-6 1.0e-6 cm ppg 10 16 rm Z 0.86 0.92 The validation was conducted by change several unknown parameters, such as annular conductivity and formation pressure, to match calculated pressure curves with field records. Values of those parameters were obtained when the match is satisfactory. 55

The matching method may also be the procedure of SCP test interpretation. The results are shown in Table 3.2. Table 3. 2 Matched Results Well 23 Well 24 KA md-ft2 0.11 0.21 pf psia 6076 6362
3.3.1 Analysis of SCP in Well 23

Well 23 was a gas production well. From the pressure record, only intermediate Casing (Fig. 3.6) had SCP problem. Casing pressure raised from 200 psi to 1600 psi and still had the increasing trend after 8 months.

738

Drive Casing 26 Conductor Casing Surface Casing 16 65# H-40 STC

1332

4310

Intermediate Casing 10 3/4 45.5# K-55 STC Production Casing 7 29# 55# N-80 LTC

11196

Fig.3. 6 Schematic of well 23 Using the model, it was predicted that the casing pressure would stabilize at about 1800 psi in 25 months (Fig. 3.7). In this case, operators were not sure about two data cement permeability and formation pressure. Cement permeability can be obtained from the matched annular conductivity by dividing it by annular cross-area. Laboratory measurements have shown that a well-cured cement typically has a permeability on the order of 0.001 md, with a 56

pore size below 2 mm and a porosity around 35% (Schlumberger Education Service, 1989). The matched cement permeability (0.35 md) is much higher than the lab value, indicating a leakage in cement. As for the formation pressure, it controls the final level the buildup pressure can reach. Only for the pressure around 6600 psia, can surface casing pressure reach at 1600 psi in 8 months. So this model can help operators to decide formation pressure and cementing quality.

2500 2000 1500 1000 500 0 0 5 10 15 Time (month) 20 25

kA = 0.11 md - ft pf = 6076 psia
2

Casing Pressure (psi)

Theory data Field data

Fig.3. 7 Pressure prediction of well 23

3.3.2 Analysis of SCP in Well 24

Before casing pressure builds up, pressure of intermediate casing in Well 24 was bled frequently. After each bleed-down, heavier mud was pumped into the annulus. Operators recorded the volume and weight of bled and pumped mud. During this buildup period, no mud was bled or pumped. From the observation, only intermediate casing (Fig. 3.8) had SCP problem. The casing pressure reached at about 1000 psia in one month. 57

582 1061 4776

Drive Pipe 26 Conductor Casing 20 94# H-40 Surface Casing 16 75# K-55

6433

Intermediate Casing 10 3/4 45.5# L-80 Production Casing 7 5/8 33# N-80

9804

Fig.3. 8 Schematic of well 24 Because of certain critical data were not collected, assumptions were made from collected records. Nevertheless, reasonable results were obtained by dividing parameters into two groups, i. e., the one controlling early SCP and the one controlling the late SCP (Fig.3.9). For the SCP in early stage, mud compressibility, gas chamber size and annular conductivity play important roles. For late period, SCP was mainly controlled by formation pressure and mud density.
1200 1000 Casing Pressure (psia) 800 600 400
Calculated Data kA = 0.21 md - ft pf = 6362 psia
2

200 0 0 5 10 15 20 Time (day)

Field Data

25

30

35

Fig.3. 9 Pressure match of well 24

58

CHAPTER 4. MATHEMATICAL MODEL OF SCP TEST

Previously developed models are all for describing SCP buildup. To completely analyze SCP diagnostic test, a mathematical model including SCP bleed-down and buildup was developed. Well geometry for this model is similar with that of previous model. Gas sequentially flows through two mediums (cement and gas-cutting mud). Finally it either accumulates at the gas chamber when the casing is closed, or evacuates from the chamber when needle valve is open. In the following sections, mechanism of gas migration in each medium will be discussed thoroughly.

Fig.4. 1 Gas cutting mud above the cement 4.1 Gas Flow in Cement The liquid column is connected with the cement at its bottom. For developing a liquid column/cement simulator, a solution for linear gas flow in cement, which related the interface pressure to the cement properties, interface flow rate, gas-source formation 59

pressure and elapsed time (Appendix C), was derived. With a constant flow rate qn during the n-th period, the pseudopressure at interface can be obtained as: m( p wf ) = m( p f ) - Dq j M (t n - t j -1 )
n

[ + (- 1)
n

4 m( p f ) - m( p c ) p

j =1

2n + 1 2 1 exp pa t n - L 2 n =0 2n + 1

(4. 1)

where
Dq j = q j - q j -1

and

M (t ) =

(2n + 1)p 16qmL 1 p 2 exp - 2 8 kAp 2 n = 0 (2 n + 1) 2L

Real interface pressure can be converted from the table of p vs. m(p) by any interpolation formula. Therefore, for a given time and flow rate, Eq.4.1 is to give the interface pressure pwf.
4.2 Gas Migration in Mud Column

Gas migration in liquid column can be modeled as dispersed two-phase flow. In dispersed two-phase flow, the dispersed phase is made up by segregated, individual particles (gas bubbles) in the midst of a continuous (liquid) phase (Cheremisinoff, 1986). The continuous phase is characterized by the fact that any arbitrary point lying in the space occupied by this particular phase can be connected with any other point in this space by means of a continuous line in such a way that all points of this line are also within this same space. On the other hand, any particle of the dispersed phase has a closed interface or boundary (almost) completely surrounded by the continuous phase. Now, it is impossible to connect a point in some particle with a point in another particle
60

without crossing the boundaries of these particles. Generally, in doing so one will pass through the continuous phase as well. The swarm of dispersed-phase particles moves through or is suspended more or less freely in the continuous phase.
Modeling Techniques Drift Flux Model. Many different modeling techniques

have been used for two-phase flows. The most common techniques are the homogeneous, drift, and two-fluid models. The two-fluid model is the most general of the three approaches. Partial differential equations for conservation of mass, momentum and energy are written for each phase. Constitutive relationships are needed for the fluid properties and to specify interaction between two phases. The two-fluid model is complex with six conservation equations (three for each phase) and a number of problematic interfacial relationships. While the two-fluid model may be appealing from a fundamentalist view-point, the complexity of the formulation and specification of the interfacial conditions make the full two-fluid model difficult to use. If the motion of the two phases is strongly coupled, as is usually the case with dispersed flows, a major simplification to the two-fluid model can be made. Instead of writing two momentum conservation equations (one for each phase), a single momentum equation can be written for the mixture as a whole resulting in the drift-flux model. The relative velocity between the phases is specified through constitutive relationships, or algebraic relationships. Most of the troublesome interfacial equations in the two-fluid model are eliminated in the approach. The idea of the drift-flux model is to concentrate on the mixture as a whole rather than the individual phase. The drift-flux model can be further simplified to give the homogeneous model. The homogeneous model assumes that each phase flows at the same velocity, or that the

61

relative velocity between the two phase is zero. The same number of conservation equations is required for the homogeneous model as for the drift-flux model. The advantage in this approach is the simplification of the constitutive relationship for the relative velocity between the two phases. In general, the applicability of the homogeneous model is limited to high mass flux situation where the relative velocity is a small fraction of the individual phase velocity. The choice of any particular model depends on the problem being considered. According to Ishii (1980), the two-fluid model is most appropriate if the two phases are weakly coupled or if the dynamic interactions between the phases is important. The driftflux model is appropriate if the phases are strongly coupled and for studies of system dynamics and instabilities. The homogeneous model is further limited to high mass flux situation where the phase velocities are approximately equal. The drift-flux model is based upon the velocity difference between the phases or between a phase and the average volumetric velocity of the mixture. The first complete presentation of the drift-flux concepts is generally considered to be by Zuber and Findlay (1965). Zuber and Findlay presented the general theory for the drift-flux approach with nonuniform velocity and void profiles applicable to a gas-liquid system. Many of their expressions for the drift flux and concentration parameters are still used today.
Conservation Equations. The most general drift-flux model consists of five

conservation equations. For most engineering problems, however, the general conservation equations can be greatly simplified. Assumptions are 1. Equal phase pressure. 2. Uniform phase densities normal to the flow direction.

62

3. Temperature is known. 4. In thermodynamic equilibrium. Due to the assumptions, the one-dimensional two-equation drift-flux model is summarized in the following: Continuity equation for dispersed phase (gas)
(ar g ) t + (ar g v g ) = 0 z

(4. 2)

Mixture momentum equation

2 f ar L r g 2 p v gm rm vm + rm vm = - - rm g - m rm vm vm - 2d h z z z t H L rm

(4. 3)

where, d h = d o - d i for flow in annuli. For the flow in pipes, d h = d (pipe diameter).

a = void fraction, H L = liquid holdup. In the case of only gas and liquid present,
a + HL =1 The mixture velocity based on the center of mass is given by 1 A r m v m dA ar g v g + H L r L v L A vm = = 1 rm r m dA A A (4. 4)

(4. 5)

The mean drift velocity is the velocity of the gas relative to the volumetric flux of the two-phase mixture, or
v gm = v g qL + qL = v g - vm A

(4. 6)

Transforming the equation based on the center of mass velocity results in v gm = H L v g - v L =

v Sg
a

- vm

(4. 7)

63

Using the fact that

v g = vs + Co vm

(4. 8)

the mean drift velocity can be expressed as

v gm = v s + (C 0 - 1)v m (4. 9)

The v gm term on the LHS represents the total mean drift velocity that consists of the v s term, or the average of the local drift/slip velocity, plus, the (C 0 - 1)v m term, which represents the convective effects. The v s term is the slip velocity, which is caused by buoyancy forces. Since the two phases have different densities, the lighter phase will rise upward relative to the heavier phase. The convective transport of the dispersed (gas) phase is due to differences in the velocity and void profiles. For example, in bubble flow, the gas phase tends to migrate toward the center of a vertical pipe. In this case, the higher velocity and void fraction occur in the center of the pipe, so more gas is convectively transported than with a uniform distribution. Thus, the mean drift velocity, v gm , is the sum of the average buoyancy effect and the convective transport effect due to nonuniform profiles. In addition to the conservation equations, constitutive relationships are needed to complete the equation set. The mean drift velocity, v gm , and the fluid fraction factor, f m must be algebraically specified. Many of these constitutive relationships are a function of the two-phase flow pattern.
Mean Drift Velocity. As mentioned previously, it can be safely assumed that two-

phase flows happened in SCP problem are in bubble and slug flow patterns. The general form of v gm used for the bubble and slug flow pattern is Eq.4.8. The value of the

64

distribution factor C 0 was selected using Hasan and Kabir (1988) method (Eq.2.7) and hydraulic diameter concept was used (Eq.2.8). In addition to the value of C 0 , v s must be specified. In general, the value of v s is a function of the flow pattern. Bubble Flow. According to Hasan and Kabir (1992), the bubble terminal velocity can be calculated using Harmathys equatioin (Eq.2.15). When considering the effect of bubble swarm, Caetano (1992) recommended Eq.2.16. Slug Flow. For vertical upward two-phase flow in annuli, no obvious difference in calculated Taylor bubble-rise velocities was observed by using mentioned equations, except for Koederitzs. General underestimation is evident by Koederitz. Hasan and Kabir (1992) method (Eq.2.28) was used in present model.
Friction Factor. Using apparent Newtonian concept (Metzner and Reed, 1955) and

Reynolds number for mixture, equations for flow in Newtonian fluid can be used to calculate Fanning friction factor. For laminar flow, Bird (1960) model (Eq.2.46) is used in concentric annuli. While in eccentric annuli, method (Eq.2.47) presented by Snyder and Goldstein (1965) was used. For turbulent flow, Cunn and Darling (1963) method was used (Eq.2.52 and
Eq.2.53). 4.2.1 Numerical Procedure

It is generally not possible to obtain analytical solutions for most practical problems in transient two-phase flow. Numerical methods based upon finite-difference concepts provide an alternate and powerful solution approach. The pressure-velocity or P-V methods (Harlow and Amsden 1971, and Patankar 1980) originally developed for single65

phase flows have had an important role in the development of computational methods for two-phase flow. Patankar (1980) provides a very good presentation of the P-V method for steadystate single-phase flow. Of particular significance to transient computational fluid mechanics is the implicit continuous Eulerian (ICE) method developed by Harlow and Amsden (1971). The ICE method for single-phase flow is able to consider flow speed ranging from imcompressibe subsonic to compressible supersonic. The basic idea of their method has had an important impact on the development of computational methods for transient two-phase flow. This is because the mixture two-phase flow equations are similar in form to the equations for compressible single-phase flow. They are complicated by the additional nonlinear terms for the mixture equation of state and the relative velocity of the liquid and gas phase. The two equations to be solved are
(ar g ) (ar g v g ) + =0 t z
( r m v m ) rv m + t z

(4. 10)

) + P + F + G + R
z

=0

(4. 11)

where
v m = (q g + q L ) / A = v Sg + v SL F= f 2 r m vm 2d h

G = rm g RM =
(ar g )(H L r L ) 2 vs rm z

66

Phase segregation after the bleed-down is a slow and gravity dominated process. Therefore, it is assumed that the relative velocity (or acceleration) term is negligible. Under these assumptions, the mixture momentum equation can be simplified as:
P + rm g + F = 0 z

(4. 12)

The finite difference solution represents the flow channel by a series of connected control volumes or computational cells. The basic concept can best be shown by example. Fig.4.2 shows a one-dimensional channel that has been subdivided into computational cells. The cell location is denoted by the index i and the cell center is denoted by the physical distance z i . Each cell has a length Dz i . The index notation i + 1 2 indicates the boundary location z i + Dz i 2 between i and
i + 1 . Similarly, i - 1 2 is the location z i - Dz i 2 between i - 1 and i.

The dashed cells at the end of the channel are used to define boundary conditions.

i-1/2 i+1/2 i -1 i i+1

N-2 N-1

DZ i

Zi
Fig.4. 2 Computational cells for one-dimensional channel It has become customary for most contemporary computational methods to use a staggered placement of variables on the computational cell as shown in Fig.4.3. The convention is to place scalars such as density, void-fraction, and temperature at cell centers. The variables are subscripted by i to denote the cell-centered location. In the conventional application of staggered grids, the pressure is also defined in the center of 67

each cell. Since an approximate momentum equation was used in present model, it is convenient to define pressure at cell faces. The velocities and mass fluxes are placed at cell faces and are denoted by full and one-half indices to identify the cell face. For example, in Fig.4.3 the velocity out the right side of the cell is vi +1 2 . For computer programming the 1/2 is not used and integer values are used by shifting the velocity index by 1/2.
i -1/2 i i +1/2

( rV )
p v
r*

( rV )

r*

Fig.4. 3 Variable placement on mass cell Finite difference analogs are used to approximate the time and space derivatives of the differential equations. They can be derived directly for the control volumes by using the integral balance laws for a control volume (Patankar 1980). They can also be derived by formal mathematical statements involving Taylors series (Forsythe and Wasow 1967). The finite difference analog for the mass equation is written as
D (ar g ) Dt + D (ar g v g ) Dz =0

(4. 13)

The specific definitions of the derivatives for a computational cell are:

D (ar g ) Dt =

(ar )
=

n +1 g i

- (ar g )i

Dt

D (ar g v g ) Dz

(ar v )
g

g i +1 2

- (ar g v g )i -1 2

Dz i

68

The superscripts n and n+1 represent time levels t and t+1, respectively. Time levels are assigned to the spatial derivatives according to the level of implicitness, presented later in the discussion. The gas mass flux (r g v g ) is defined as

(r v ) = r
g g

*
g

vg

where r g is the gas density at the cell boundary. A variety of methods (Harlow and Amsden 1971, and Liles and Reed 1978) have been suggested to define density at the cell boundary. An attractive form is a weighted donor-cell definition (Liles and Reed 1978) given by

(r )

* g i +1 2

1+ b (r g )i + 1 - b (r g )i+1 2 2

where b can vary between 1 and +1. Choice of b = 0 produces a mathematically more accurate but less stable central difference. If b is defined by
b = sign (v g )i +1 2

the full-donor differencing is obtained where density is assigned from the upstream density. The same logic applies to the definition of all scalar required at cell boundary. As mentioned above, a simplified momentum equation was used. It is convenient to define pressure at cell faces and estimate the pressure at the center by an arithmetic average, i.e., pi = ( p i +1 2 + p i -1 2 ) 2 . The finite difference analog of the momentum equation can be written as
DP +G+ F = 0 Dz

(4. 14)

The pressure gradient is defined as 69

DP Pi +1 2 - Pi -1 2 = Dz i Dz i

The friction term for turbulent flow is defined as

f 2 Fi = r v m m 2d h i

The gravity term is defined as Gi = r m g z

where the average density is calculate by the pressure estimated at the center,

(r m )i

= r m ( pi )

The previous discussion has presented finite-difference equations without specific reference to the time level of the spatial difference terms or the source terms. The level of implicitness (or explicitmess) is concerned with making those time specifications. This has an important affect on the time-step limits and stability of the numerical solution.
Explicit. The fully explicit equations are written as:

D (ar g ) Dt

D (ar g v g ) Dz

=0

(4. 15)

DP n + F n + Gn = 0 Dz

(4. 16)

where all source and derivative terms are at time n. Explicit methods are simple because the temporal derivatives are evaluated directly to advance the solution from n to n+1. The explicit treatment of the mass and momentum equation limits the time step to be on the order of a sonic transit time through the most limiting computational cell. This limit is usually severe enough that the explicit method is not used for most practical two-phase flow numerical solutions.

70

Implicit. The fully implicit equations can be written in the form

D (ar g ) Dt

D (ar g v g ) Dz

n +1

=0

(4. 17)

DP n+1 + G n+1 + F n +1 = 0 Dz

(4. 18)

where all spatial derivatives and source terms are now at time n+1. Fully implicit equations can substantially reduce or eliminate time step restrictions at the expense of a simultaneous solution of all variables. For an arbitrary flow direction this can require a rather complex solution. The complexity is great enough that fully implicit methods are not available for transient two-phase flow.
Semi-implicit. Between the extremes of explicit and implicit are the semi-implicit

equations. An important semi-implicit selection is:

D (ar g ) Dt + D ar g

* n

)v

n +1 g

Dz

)= 0

(4. 19)

DP n +1 + Gn + F n = 0 Dz

(4. 20)

Note that the mass equation is implicit in velocity and the momentum equation is implicit in pressure. The time step limit is on the order of the fluid transit time through a computational cell rather than the sonic transit time through a cell as for the fully explicit equations. This usually results in time steps that are several orders of magnitude larger than the time step for a typical application of the explicit solution method. In this study, semi-implicit method was use to calculate the void fraction in each cell. Once the volume of gas in each cell is estimated for a particular time step, the gas volume at the wellhead can be calculated from the annular volume between casings and the total volume of the liquid and gas at that instant in the rest of the annular. 71

4.2.2 Boundary Conditions

During SCP buildup, gas accumulates at annulus top with closed needle valve. While in SCP bleed-down, gas is released from the top when needle valve is open. Therefore, two different upper boundary conditions are considered.
4.2.2.1 Gas Discharg through a Needle Valve

The bleed-down procedure normally involves gas or gas-liquid flowing through a needle valve to the atmosphere. It is can be considered as single-phase gas or multiphase flow through a choke. Gas usually flows at sonic velocity and its flow rate is easily to record (Fig.4.4).
25

20

Gas Rate (MSCF/D)

15

10

0 0 5 10 15

Time (min)

Fig.4. 4 Gas rate during bleed-down While for multiphase, the critical-flow boundary changes with the in-situ gas/liquid ratio R1 (Ashford and Pierce, 1975). According to MMS policy, the volume of liquid collected and bleed-down time should be recorded in SCP diagnostic tests. Thus, liquid flow rates can be determined. If gas flow rate cannot be measured directly, it can be obtained from the following iterative procedure. 1. Initial guess that yc* = 0.5. 72

2. If y < yc*, critical flow exists. Calculate the gas flow rate qgsc using Eq.2.72 (Gilbert, 1954). If not, subcritical flow exists. Using TUFFP model (Eq.2.65) to calculate the gas flow rate. 3. Using the calculated gas flow rate in step 2, calculate gas/liquid ratio R1. Then, an iterative procedure (Eq.2.64) is used to determine value of pressure ratio yc. Let yc* = yc. 4. Repeat step 2 to 3 until there is no change in flow condition.
4.2.2.2 Gas Accumulation at Gas Chamber.

When developing a model to simulate the wellbore phase redistribution in well testing, Hasan and Kabir (1992) assumed that a gas chamber formed instantaneously after the shut-in. Same gas chamber was also assumed in this model. For the purpose of computation, this gas chamber is treated separately from all the other cells. The chamber is completely filled with gas (no liquid). When the wellhead is closed and gas has the tendency to migrate upward, it doesnot lose gas to any other cell but receive gas from the cell immediately below it. The volume of this wellhead gas chamber changes with time. The chamber receives gas from the cell below and gas (and mud) lost in bleed-down tending to increase its volume. On the other hand, the expansion of gas in the rest of well tends to decrease the gas chamber volume. The net effect of these two opposing processes may be a net increase or decrease in the wellhead gas chamber volume. The wellhead gas chamber volume, Vwh, at any time n+1, is related to its volume at the earlier time step, n, in the following manner,
n +1 n Vwh = Vwh + (V g ) j + (VL ) j - (V g ) j - (VL ) j - (q g )0 Dt + (q m ) N Dt (4. 21) n n n +1 n +1 n +1 n +1 j =1 j =1 j =1 j =1 N -1 N N -1 N

73

where N is the number of equal-sized cell in well, the two summation terms for Vg are the change in wellhead gas chamber volume due to the net expansion of the gas in the rest of the liquid column. The summation of VL indicates the change in the volume of liquid due to increased liquid column pressure during this time step. The second to last term in RHS is the volume decrease caused by gas flow from cement below the liquid column. The last term is the volume increase caused by gas flow out from the annulus. During casing pressure buildup, the term is zero, since the casing is shut in. The wellhead pressure at any time n+1, is related to the pressure at the previous time step, n, through the gas law, p
n +1 wh n n n +1 p wh n n +1Vwh z = n n +1 n n Vwh z

(4. 22)

Once the volume of gas chamber is estimated by Eq.4.21, the wellhead pressure can be calculated from Eq.4.22. This step, in turn, allows calculation of interface pressure from the wellhead pressure by Eq.4.20. The entire calculation procedure is iterative.
4.2.2.3 Boundary Condition at Interface between Cement and Mud Column

The simplest method to specify boundary conditions is to use a zero thickness phantom boundary cell. Fig.4.5 shows an example of boundary cells and boundary conditions specified for pressure and mass flux. The velocity boundary condition is assigned by defining the velocity through a cell face and appropriately modifying the pressure gradient at the boundary. The velocity can be an arbitrarily specified value, including zero for a non-flow boundary. In present model, a velocity boundary condition is set at the interface in each time step.

74

P GIVEN; V CALCULATED V P V

V GIVEN; P CALCULATED V P V

V = 0; P CALCULATED CALCULATED PRESSURE SPECIFIED PRESSURE Fig.4. 5 Examples of boundary conditions in boundary cells
4.2.3 Solution Algorithm

The semi-implicit equations used for the solution are Eq.4.19 and Eq.4.20. The iterative method involves a simultaneous solution of the mass and momentum equation for pressure and velocity in the rest of liquid column. With an assumed interface pressure, density can be calculated from the equation of state. Gas velocity and void fraction can be calculated from the mass equation at cells in the rest of liquid column. Then the volume and pressure of wellhead gas chamber can be calculated through gas law. By adding the hydrostatic pressure of previous time step and current frictional pressure to the wellhead pressure, another interface pressure can be obtain. If the difference between those two interface pressures is small enough, the iteration is stopped. The overall procedure is shown in Fig.4.6.
75

Define the gas flow rate at the interface, qc ~ Guess a value of interface pressure pws i=1
* Guess a value of pressure pi +1 2

Calculate the average pressure pi and solve EOS for density

Solve mass Eq. for void fraction ai

Calculate the mixture density rm and solve momentum Eq. for pressure p i + 1 2

pi +1 2 - pi*+1 2 < e N pi*+1 2 = 0.5( pi*+1 2 + pi +1 2 )

Y i=i+1 Calculate the volume of gas chamber and pressure at wellhead pwh Calculate the interface pressure pws ~ pws = 0.5( ~ pws + pws )

~ pws - pws < e

Next time step

Fig.4. 6 Solution diagram for two-phase flow in liquid column in one time step The steps for one iteration cycle with the specific interface gas flow rate qc are as follows: 1. Guess a tentative pressure ~ p ws 1) Guess a value of pressure pi*+1 2 . 2) Calculate the average pressure for r g , r L .

76

3) Solve the mass Eq.4.19 and momentum Eq.4.20 for a and v g . 4) Calculate the pressure pi +1 2 . 5) Check if converged. If so, go to next cell. If not, guess a new value of pressure pi*+1 2 and go back to step 2. 6) Repeat step 2 to step 6, until calculations at all cells below gas chamber are converged. 2. Calculate Vwh using Eq.4.21 and pwh using Eq.4.22. 3. Calculate pws using momentum Eq.4.20.
p ws - p ws < e . 4. Repeat cycle until converged, i.e., ~
4.3 Coupling Gas Flows in Liquid Column and Cement

The coupling procedure is shown in Fig. 4.7. Given that each system (cement and liquid column) has been set up from input, the solution is started from a defined set of initial and boundary conditions. Difference between pressure bleed-down and buildup is revealed at upper boundary, with positive flow rates for bleed-down and zero flow rates for buildup. With a specific gas flow rate at interface qc at each time step, hybrid approach is used to couple numerical solutions in liquid column and analytical solutions in cement. Coupling criteria is set to be that pressure above (pws) the interface is not less than pressure below (pwf).
Coupling Criteria: pws = pwf for qc > 0 or pws > pwf for qc = 0 Upper Boundary Condition: qg and qL Needle Valve
Initial Condition: Fg (z, 0) and p (z,0) Numerical Solution: Fg (z,t) and pws

Initial Condition: p (z, 0) Analytical Solution: pwf Interface Cement

Lower Boundary Condition: pf = constant Formation

Liquid Column

Fig.4. 7 Coupling procedure 77

When coupling criteria at any time step is satisfied, casing pressure is equal to pressure in gas chamber and it can be obtained by subtracting pressure losses in two systems from formation pressure:
p csg = p f - Dp c - (Dp m + Dp f

(4. 23)

where, p f is formation pressure, Dpc is the pressure loss in cement, Dp m is the hydrostatic pressure of gas-liquid mixture in liquid column and Dp f is friction pressure loss caused by swarm of bubbles moving against annular walls. Detailed algorithms for coupling will be discussed below. First procedure needed to simulate is the bleed-down because SCP diagnostic test begins with it. At initial time, the surface pressure, size of gas chamber and the gas concentration in liquid column are known and pressure gradient can be calculated. Constant gradient was assumed throughout the entire column (Walker, 1936 and McCoy et al., 1988). Assuming zero flow rate at the interface between cement and liquid column, the pressure and gas distribution in liquid column as well as pressure at the interface can be calculated. An assumption was made that cement is filled with gas whose density is negligible. Therefore, the pressure is uniform and equal to interface pressure in entire cement, except at the point of gas-source formation, where it is equal to constant formation pressure and is higher than the interface pressure (Fig.4.8). From bleed-down pressure history, gas (and liquid) rate at each recorded time interval was determined. With the initial condition and two boundary conditions (known flow rates at wellhead and constant pressure at gas-source formation), pressure distribution in liquid column is calculated by the iterative method described (Fig.4.6).

78

Pressure distribution in cement is determined by the solution of linear gas flow (Eq.4.1). Iteration stops until the flow rate and pressure at either side of interface are equal.
Pressure (psia)
0 0 1000 2000 3000 4000 5000 6000

-1000

-2000

Depth (ft)

End of bleed-down

Beginning of bleed-down

-3000 Interface between cement and mud column

-4000

-5000

Gas-source formation

-6000

Fig.4. 8 Pressure distribution during bleed-down When a good match for bleed-down is obtained, initial conditions at the instant of shut-in is known and the calculation of variables during pressure buildup can be carried on with the different upper boundary condition (zero flow-rate at casing head). With initially assumed zero gas flow-rate (qc) at interface, the pressure distributions in liquid column and cement are calculated respectively. If there is a pressure difference favoring gas flow through the interface, recalculate the pressure distributions using guessed gas flow rate (qc*) at interface. Repeat the cycle until pressures at the either side of interface are equal for a guessed gas flow rate. Time is then incremented, boundary conditions defined, and the process is repeated until the desired time is reached. During the whole process, assumption is made that mud does not flow back to cement because of its high viscosity and low relative permeability to cement and only gas flow through the interface. The force that drives gas flowing through the interface is the 79

pressure difference between below (pwf) and above (pws) the interface. The pressure increase at cement top is caused by the continuous gas charge from high-pressure gassource formation. The interface gas flow rate keeps zero until a positive (upward) pressure difference appears at the interface. The iterative procedure discussed is shown in
Fig.4.9.
Define initial conditions

Set boundary condition qc = 0 In one time, step, determine the pressure above the interface pws and the pressure below pwf No

pwf > pws Yes

Assume an interface gas rate qc* Recalculate the pressure above pws and the pressure below pwf Update gas rate qc* No |pwf - pws| <e Yes t = t + Dt Yes Next Dt? No Stop

Fig.4. 9 Solution diagram for gas flow in liquid column and cement

80

CHAPTER 5. THEORETICAL ANALYSIS OF SCP TESTING

The mathematical model was applied to provide theoretical explanations for different SCP typical patterns, engineering scenarios in SCP tests and selected field records. Sensitivity analysis of five parameters on pressure bleed-down pressure and buildup model was also conducted using theoretical results computed by the model. 5.1 Understanding of SCP Typical Patterns Five patterns for pressure bleed-down and buildup were concluded from the field data. Using the mathematical model, reasons for those characteristic behaviors are revealed. 5.1.1 Bleed-Down Patterns Instant Bleed-Down Pattern. As mentioned before, this pattern is characterized with rapid casing pressure drop and liquid removal. A theoretical pressure curve reflecting the feature was drawn using the model. Reasons for the pattern was found to be low gas supply above cement, such as small gas chamber (1.2-ft long), low initial gas concentration in mud (0.08%), and slow gas supply in cement because of low annular conductivity (1.1 md-ft2). With a wide opened needle valve, mud is removed with gas. It should be pointed out that when the theoretical curve was plot in shorter time interval than the records, pressure drop was not linear (Fig.5.1). The early pressure drop can be explained as gas evacuation from the gas chamber. This period is very short because of the small gas chamber. Then loss of liquid from annuli reduces hydrostatic pressure in liquid column and causes high pressure-difference at the interface, prompting gas supply from cement to liquid column. This will compensate the quick early pressure drop and less pressure decrease with time should be expected. This result also proved that instant

81

bleed-down would not bled casing pressure to zero as long as there is a leaking cement and a high-pressure gas source formation. The first discontinuity of pressure drop trend indicates the transition of flow mechanism and may give important information on the gas cap size. The ratio of pressure drop to initial pressure (

pi - p FD ) represents the pi

interaction between gas chamber and gas in liquid. When the ratio is equal to 1, it means that all bleed-down is mainly evacuation of gas chamber with no gas in liquid. If the ratio is close to zero, gas chamber is very small and the bleed-down is dominated by two-phase flow through the needle valve caused by expansion of gas in the liquid column.
800 Lg = 1.2 ft Fg = 0.008 Casing Pressure (psia) 600 Field Data Early Pressure Drop

400

200

First Discontinuity Point

0 0

t FD

5 Time (sec)

10

15

Fig.5. 1 Theoretically instant bleed-down pattern Long Bleed-Down Pattern. A long bleed-down pattern shows non-linearly pressure decrease and almost no liquid removal. In present model, the pattern (Fig.5.2) was simulated with restrained needle valve and high gas supply, such as 23-ft long gas chamber and 0.25% gas in mud initially. It is found that first slow decrease is due to the manipulation of needle valve and last stabilization is caused by steady-state flow in the whole system, i.e., gas supply balances the gas bled off. This pattern is actually a special 82

case of instant bleed-down with restricted opening of needle valve and without liquid removal from annuli. Therefore, casing pressure decrease curve will flat out at late time but could not reach to zero. In this situation, it is very critical to record any change in opening of needle valve, since the size of choke is very important to calculate the gas flow rate from the pressure history and further the shape of the pressure curve.
1600

Casing Pressure (psia)

1200

Lg = 23 ft Fg = 0.0025

800

Non-linear
400

Steady-state Flow

4 Time (min)

12

Fig.5. 2 Theoretically long bleed-down pattern 5.1.2 Buildup Patterns Normal Buildup Pattern. In this kind of pattern, casing pressure completes the whole cycle of buildup, including early-time increase, transition and late-time stabilization. Casing pressure increase at early time is caused by gas migrating and releasing from mud. The magnitude of increase is controlled by the amount of gas trapped in mud. If there is a lot of gas in mud, the increase is prominent. On the contrary, the increase is unnoticeable if there is little gas in mud. The duration of this early time behavior is controlled by gas rising velocity in mud. The boundary between early-time behavior and transition is when the first swarm of gas from cement reaches the gas chamber. 83

Therefore, if the length of liquid column is known, the gas rising velocity can be determined by the duration of early time SCP buildup. Transitional behavior is mainly controlled by the gas supply from cement. The magnitude of casing pressure increase is affected by the amount of the gas supply. If gas flow rate at the cement top is high because of high annular conductivity or/and high gas supplying pressure at gas formation, casing pressure increase is high, and vise versa. The transition ends when the pressure at liquid column bottom is high enough to balance the pressure at cement top causing no gas flow at the interface. This is also the beginning of late-time stabilization (Fig.5.3). From there on, pressure will continually increase as residual gas in liquid migrates upwards. But this increase is trivial because only small amount of gas was left in the liquid column at the end of transition.
1400 1200

Casing Pressure (psia)

1000

Early Time

Transition

Late Time

800 600 400 200 0

10

15

20

25

30

35

40

Time (hour)

Fig.5. 3 Theoretically normal buildup pattern S-shaped Buildup Pattern. From above discussion, the s-shaped buildup pattern is the special case of normal one. Unlike the normal pattern, no gas is in liquid column at early stage of buildup (Fg = 0). Increase in pressure is purely caused by compression of mud by gas flowing from cement. Therefore, the increase is almost invisible. 84

When the first swarm of gas from cement migrates to the annulus top, casing pressure begins to behave as transition. Pressure keeps rising as gas supply continues. When the pressure reach a balanced condition in the system, casing pressure stabilizes at certain level and the buildup is at late stage. Theoretical curve for the pattern is shown in Fig.5.4.

1400 1200

Normal Pattern

Casing Pressure (psia)

1000 800

S-shaped Pattern
600 400 200 0 0 5 10 15 20 25

Time (hour)

Fig.5. 4 Theoretically s-shaped buildup pattern

Long Buildup Pattern. After bleed-down, only small amount of gas is left in the gas chamber. Therefore, any gas charge from the liquid column will result in obvious pressure increase in early time. If formation pressure is high enough to keep a pressure difference at interface, constant gas flow from cement will produce slow but continual pressure increase. Pressure can be expected to stabilize at certain value in a very long time (more than 24 hours) and the stabilized pressure might be dangerously high.

85

1400

Casing Pressure (psia)

1200

Early Time

Transition

1000

800

600 0 5 10 15 20 25

Time (Hour)

Fig.5. 5 Theoretical long buildup pattern 5.2 Understanding of SCP Testing Scenarios Engineering scenarios are combinations of different conditions of two mediums; liquid column and cement. These different conditions manifest themselves in the present model as different levels of parameters for calculating the theoretical pressure. For example, low initial gas concentration in mud represents thin mud. High annular conductivity stands for poor cement bondetc. By setting different value to certain parameters in this model, explanation and prediction of SCP behavior at different engineering scenarios can be made. 5.2.1 Thin Mud Combined with Poor Cement Bond. Thin mud traps no gas or very little gas inside it, i.e., initial gas concentration in mud is very low. Therefore, the bleed-down procedure is virtually unloading of the top gas chamber accumulated at previous buildup. After the bleed-down, casing pressure remains relatively constant until the first swarm of gas reaches the casing head. On the other hand, poor cement bond provides an easy path for gas flowing through it. A period of obvious pressure increase following a slow early-time 86

buildup is expected. With this combination of mud and cement bond, the duration of transitional zone is controlled by the size of gas chamber when the bleed-down begins. If the chamber is large, casing pressure increase is relatively slow, leading to either very late pressure stabilization or no stabilization at the end of testing time (usually 24 hours). The mathematical model can clearly illustrate the situation (Fig.5.6).
1400 1200 Casing Pressure (psia) 1000 800 600 400 200 0 0 5 Vwh = 900 gal (SC) Vwh = 90 gal (SC) 10 15 Time (hour) 20 25

Fig.5. 6 Theoretical casing pressure behavior for thin mud and poor cement bond Combined With Good Cement Bond. In this case, similar casing pressure behavior during bleed-down and early-time buildup can be expected. However, with good cement bond, gas flow rate is very low in the cement. Therefore casing pressure increases very slowly after the early-time buildup, leading to a long transitional time before casing pressure stabilizes at certain value. It is quite possible that no late-time behavior shows up during the testing time. In this situation, gas flow rate in cement is the dominant factor controlling casing pressure buildup. Initial gas chamber volume only affects final pressure bled down and does not control the transitional time. With the same properties except for casing conductivity, the theoretical casing pressure curve was calculated using current model (Fig.5.7). 87

800

Casing Pressure (psia)

600

400

200 Vwh = 900 gal (SC) Vwh = 90 gal (SC) 0 5 10 15 20 25

0 Time (hour)

Fig.5. 7 Theoretical casing pressure behavior for thin mud and good cement bond 5.2.2 Thick Mud Combined with Poor Cement Bond. Thick mud traps more gas than thin mud. The average gas concentration in mud is high during the bleed-down and buildup. Gas migration in mud is the dominant factor controlling whole casing pressure buildup. In the case of relatively low initial gas concentration, casing pressure stabilization follows a rapid early-time increase and a transitional zone. While for the case of high gas concentration, no obvious transitional zone appears. After the gas trapped in mud reaches casing head, the hydrostatic pressure of mud plus the surface pressure is high enough to prevent further gas flowing from cement. The curves calculated from the mathematical model clearly illustrate these two situations (Fig.5.8). Combine with Good Cement Bond. Very little gas can flow through a good cement bond. Therefore, the dominant factor is gas migration in mud. With a weak gas supply from cement, transitional time is too short to be seen. Casing pressure will remain constant after all the gas previously trapped in mud reaches the casing head. But that is not the end of the story for the case of low initial gas concentration. As long as the 88

hydrostatic pressure of mud plus surface pressure is lower than the pressure at the top of cement, gas will flow through the interface and manifests into continuously increasing pressure. Fig.5.9 shows casing pressure responses calculated by the present model for different initial gas concentration.

1600

Casing Pressure (psia)

1200

800

400 Fg = 0.005 Fg = 0.001 0 0 10 20 Time (hour) 30 40

Fig.5. 8 Theoretical casing pressure behavior for thick mud and poor cement bond

1600

Casing Pressure (psia)

1200

800

400 Fg = 0.005 Fg = 0.001 0 0 5 10 15 Time (hour) 20 25

Fig.5. 9 Theoretical casing pressure behavior for thick mud and good cement bond

89

5.2.3 Slightly Compressible Mud vs. Incompressible Mud In the model, the effect of mud compressibility was neglected during the bleeddown, since the system is open to atmosphere and difference in compressibility between gas and mud is large. During casing pressure buildup, the system is closed. Mud compressibility plays an important role in controlling buildup behavior. Compressible mud produces less increase in casing pressure than incompressible mud (Fig.5.10).
1600

Casing Pressure (psia)

1200

800

400 Cm = 0 1/psi Cm = 7e-6 1/psi 0 0 5 10 15 20 25 Time (hour)

Fig.5. 10 Theoretical casing pressure behavior for mud with different compressibility 5.3 Sensitivity Analysis Using the mathematical model, investigation was conducted on several factors affecting pressure bleed-down and buildup. The reason for the selection is that those factors are either hard-to-obtain or important for remediation actions. They are formationpressure, annular conductivity, mud compressibility, initial gas concentration in mud and initial size of gas chamber. The significance of those factors on calculated bleed-down and pressure buildup was tested using several statistical methods. The purpose of sensitivity analysis is first to identify controlling factors for pressure bleed-down and buildup. This will be helpful for 90

operators to collect important data for SCP testing and for making efficient remedial actions to finally eliminate the SCP. Another purpose is to investigate the sufficiency and efficiency of current requirements for SCP testing. 5.3.1 Approach Pressure Bleed-Down. To eliminate the effect of initial pressure, the ratio of pressure difference ( Dp SBD ) between initial pressure ( pi ) and stabilized pressure in bleed-down ( p SBD ) to initial pressure was selected as a function of four parameters. p i - p SBD Dp SBD = = f ( p f , k , Fg ,V gc ) pi pi

(5. 1)

where, p f is formation pressure, k is cement permeability, obtained by dividing annular conductivity by constant annular cross-area; Fg is initial gas concentration in liquid; and V gc is the volume of gas chamber. Sensitivity of pressure difference to those four factors was tested to identify the controlling parameters on pressure bleed-down. The value of Dp SBD pi indicates efficiency of the bleed-down. For Dp SBD pi = 1 , the bleed-down is complete, i.e., pressure reduces to zero. Values of Dp SBD pi close to unity represent that pressure cannot be bled down thus indicating a severe SCP problem. The first discontinuity time ( t FD ) was defined as the time when the gas cap evacuation ends (Fig.5.1). (The time indicates transition of the flow through choke from the single-phase to two-phase flow.) The gas evacuation stage is important for measuring volume of gas chamber. Thus, the gas evacuation time should be available from earlytime pressure record an important requirement for pressure monitoring system.

91

Flow Schedule for Bleed-down. Two specific needle-valve flow rate schedules have

been assumed to simulate rapid and prolonged bleed-downs. Each bleed-down begins at assumed value of initial pressure and ends at the time past pressure stabilization as shown in Table 5.1. Factors t rec Table 5. 1 Inputs for Bleed-Downs Unit Instant Bleed-Down Long Bleed-Down sec 14 720

pi

psi

695

1354

For instant bleed-down, liquid removal is emphasized with increasing liquid flow rate and decreasing gas flow rate through the needle valve (Fig.5.11). The duration of bleed-down lasts only 14 seconds to reach the stabilization in most runs. For long bleeddown, decreasing gas flow rates are assumed to simulate MMS recommended no-liquidremoval schedule (Fig.5.12). In 12 minutes, most runs of pressure bleed-down reach stabilization. Therefore, the pressure values recorded at 14 seconds and 12 minutes have been used as stabilized bleed-down pressure, p SBD , in Eq.5.1.
30000 25000 20000 15000 15000 10000 5000 0 0 5 Time (sec) 10 15 10000 5000 0 35000 Liquid Flow Rate (BBL/D) 30000 25000 20000

Gas Flow Rate (SCF/D)

Fig.5. 11 Flow schedule for instant bleed-down

92

1000

Gas Flow Rate (SCF/D)

800

600

400

200 0 2 4 6 8 10 12 14 Time (min)

Fig.5. 12 Flow schedule for long bleed-down It should be noted that imposition of only two specific flow rate schedules limits generality of the sensitivity study. Further research should modify the model to simulate more general cases without a pre-designed flow schedule.
Pressure Buildup. To eliminate the influence of bleed-down, incremental pressure

buildup, Dp1 , was used as a function of five factors.

Dp1 = p BU 1 - p BD = f ( p f , k , c m , Fg , V gc )

(5. 2)

where

p BU 1 = the casing pressure after one day (24 hours) of buildup (time limit for SCP
diagnostic test required by MMS)

p BD = the shut-in pressure at the end of bleed-down c m = mud compressibility

The purpose of sensitivity analysis was to find required duration of pressure buildup. In principle, the buildup test should be long enough to become sensitive to a parameter of interest that must be determined from the test. In case a 24-hour buildup was not 93

sufficiently long, a two-day or three-day buildup should be tested for sensitivity to that parameter. The corresponding incremental pressure buildup are defined as:
* Dp 2 = p BU 2 - p BU 1 = f ( p f , k , c m , Fg ,V gc ) * Dp3 = p BU 2 - p BU 1 = f ( p f , k , c m , Fg ,V gc )

(5. 3) (5. 4)

where, p BU 2 and p BU 3 are casing pressure buildup after two and three days, respectively.
5.3.2 Methods of Sensitivity Analysis of SCP Testing

The sensitivity analysis used multiple linear regression and two level factorial design. First, low and high levels of each factor were selected and casing pressure response was computed for at those two levels using the mathematical model. Then the F-test was used to test if there was a linear relationship between calculated pressure and those factors. Finally, a T-test was performed to identify the significance of each factor contribution to the pressure regression model. Each process will be discussed thoroughly in following sections.
Two-Level Factorial Designs. Factorial designs are widely used in experiments

involving several factors where it is necessary to investigate the joint effects of the factors on a response variable (Myers and Montgomery, 1995). A very important special case of the factorial design is that where each of the k factors of interest has only two levels. These designs are called 2k factorial designs. In two-level factorial designs, levels of the factors usually are coded to 1 and +1. And the scaling formula for a factor is
y= 2x - a - b b-a

(5. 5)

where a and b are low and high level of the factor. In this study, each factor is assigned with a reasonable range (Table 5.2) and is scaled to 1 and +1 using Eq.5.5. Other data
94

input as constants in the sensitivity study are listed in Table 5.3. The formation pressure was determined using minimum and maximum possible hydrostatic pressure of liquid in annuli at the specific depth of gas-source formation. In this study, a complete factorial design in 5 factors was used. Therefore, 25 = 32 runs are performed by calculating pressure bleed-down and buildup using all possible combination of two levels. Two-level factorial design is useful in identifying the important system properties for SCP testing. Table 5. 2 Range of Each Factor Feasible Range Unit Instant Bleed-Down Long Bleed-Down psi md-ft2 psi
-1

Factors
pf
k

5018 to 5552 1.1 to 11 0 to 7 10 -6 0 to 0.005 90 to 900

5475 to 6000 1.1 to 11 0 to 1.5 10 -5 0 to 0.009 900 to 9000

cm
Fg V gc

gal

Table 5. 3 Inputs for Simulations Properties Unit Instant Long System Bleed-Down Bleed-Down Inner diameter in 9.95 9.95 Annulus Outer diameter in 7.625 7.625 Length ft 598 597 Cement Porosity 0.01 0.01 Formation Depth ft 5447 5447 Gas Specific gravity 0.71 0.71 Consistency index eq cp 285.64 285.64 Flow behavior index 0.8 0.8 Liquid Density ppg 17.2 16 Interface tension dynes/cm 67.5 67.5 Multiple Linear Regression Models. In this study, it is assumed that (the dependent variable) casing pressure p csg is related to 5 independent variables (Table 5.2) for during buildup and bleed-down. 95

The multiple regression models is

p csg = b 0 + b 1 p f + b 2 k + b 3 c m + b 4 Fg + b 5V gc + e

(5. 6)

for pressure buildup, and

p csg = b 0 + b 1 p f + b 2 k + b 3 Fg + b 4V gc + e

(5. 7)

for the bleed-down. The parameters, b j , j = 0,1, K ,5 , are called the regression coefficients and typically are estimated by method of least square. The method chooses the b s in Eq.5.6 or Eq.5.7 so that the sum of the squares of the errors are minimized. The linear regression model is mainly used to filter the important controlling parameter but accurate description of the system may need high-order models.
Hypothesis Testing in Multiple Regression. The test for significance of regression

is a test to determine if there is a linear relationship between the response variable (calculated casing pressure) and a subset of the regressor variables (factors), for example,
p f , k , c m , Fg and V gc for pressure buildup and p f , k , V gc , Fg and t BD for bleed-down.

The appropriate hypothesis are H 0 : b1 = b 2 = L = b 5 = 0 H 1 : b j 0 for at least one j (5. 8)

Where, b j is the regression coefficient of each factor. Rejection of H0: in Eq.5.8 implies that at least one of factors contributes significantly to the model. The total sum of squares Syy is the sum of a sum of squares due to the regression, SSR, and a sum of squares due to residual (or error), SSE.
S yy = SS R + SS E

(5. 9)

96

The test procedure for H 0 : b1 = b 2 = L = b k = 0 is to compute F0 = SS R k MS R = SS R (n - k - 1) MS E (5. 10)

and to reject H0 if F0 exceeds Fa ,k ,n -k -1 . Alternatively, one could use the P-value approach to hypothesis testing and, thus, reject H0 if the P-value for the statistic F0 is less than a . The test is called an analysis of variance (ANOVA). And it is usually summarized in an ANOVA table (Table 5.3). In this study, a confidence level as 98% ( a = 0.02 ) was select, meaning which the null hypothesis is rejected if probability (Pvalue) is less than 2%. The coefficient of multiple determination R2 is defined as R2 = SS R SS = 1- E S yy S yy (5. 11)

A large value of R2 does not necessarily imply that the regression model is good one. Adding a variable to the model will always increase R2, regardless of whether the additional variable is statistically significant or not. Some regression model builders prefer to use an adjusted statistic R2 defined as
2 Radj = 1-

SS E (n - k ) n -1 2 = 1- (1 - R ) S yy (n - 1) n-k

(5. 12)

2 When R2 and Radj differ dramatically there is a good chance that non-significant

terms have been included in the model. Table 5. 4 ANOVA for Significance of Fitted Model Source of Degrees of Sum of Mean Significance of F0 Variation Freedom Squares Square F0 Regression k SSR MSR MSR/MSE P-value Error n-k-1 SSE MSE Total n-1 Syy 97

Tests on Individual Regression Coefficient. SCP is typically driven by several

physical mechanisms. Using T-test, significance of individual regression coefficient can be tested. Therefore, corresponding factors that significantly contribute to the model can be determined. The hypothesis for testing the significance of individual regression coefficient, b j , are
H0 : b j = 0 H1 : b j 0 If H 0 : b j = 0 is not rejected, it indicates that the corresponding factor can be deleted from the model. The test statistic for this hypothesis is
t0 = se(b j ) bj

(5. 13)

where bj is the estimation of b j using least square method. The null hypothesis
H 0 : b j = 0 is rejected if t 0 > ta / 2,n - k -1 . The denominator of Eq.5.13 is called the

standard error of the regression coefficient bj. Like ANOVA, this T-test can be performed using statistical software such as Microsoft Excel or SAS and the results are listed in a table.
5.3.3 Analysis of Bleed-Down Pressure

Effects of four factors on stabilized bleed-down pressure (Eq.5.1) were investigated. Using two-level of factorial design, 24 = 16 observations were respectively calculated with the mathematical model for instant and long bleed-down. The significance of the fitted model for those pressure drops was tested. P-values (1.45E-16 and 1.59E-5, respectively) for F0 were much lower than selected a (0.02), the hypothesis

98

H 0: b1 = b 2 = K = b 4 = 0 was rejected, which implied that at least one of the variables contributed significantly to the model. The estimates for individual coefficient were listed in Table 5.4 and Table 5.5. Both results of T-test showed that P-values of initial gas concentration and size of gas chamber were far less than 0.02. Therefore, they significantly contributed to the efficiency of bleed-down, i.e. the ratio Dp BD pi . Comparing the coefficient for these two parameters showed that effect of volume of gas chamber was stronger than that of initial gas concentration in liquid column. Table 5. 5 Coefficient Estimates for Instant Bleed-Down
Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Intercept 0.80 0.0 614.1 2.68E-26 0.8 0.8 pf 0.00 0.0 0.0 1 0.0 0.0 k 0.00 0.0 0.0 1 0.0 0.0 Fg -0.03 0.0 -26.0 3.18E-11 0.0 0.0 Vgc -0.14 0.0 -105.0 7.29E-18 -0.1 -0.1

Table 5. 6 Coefficient Estimates for Long Bleed-Down

Intercept pf k Fg Vgc Coefficients Standard Error t Stat P-value Lower 95% Upper 95% 0.4 0.0 18.6 1.14E-09 0.4 0.5 0.0 0.0 0.0 1 0.0 0.0 0.0 0.0 0.0 1 0.0 0.0 -0.1 0.0 -5.3 2.65E-04 -0.2 -0.1 -0.2 0.0 -8.6 3.11E-06 -0.2 -0.1

Minimum value of first discontinuity time ( t FD ) was from 2 seconds, indicating that the evacuation of gas chamber could occur in very short time. Current monitoring time interval is 14 seconds, which is too long to detect the change in pressure drop trend and obtain useful information of gas chamber.
5.3.4 Analysis for Buildup Pressure

First, sensitivity of 24-hour buildup to five factors (Eq.5.2) was investigated for both instant and long bleed-down procedure. 99

Buildup after Instant Bleed-Down. ANOVA for regression (Table 5.6) showed the

P-value (5.93E-7) was far less than 0.02, i.e., at least one factor contributes significantly to the model. Table 5. 7 ANOVA for Buildup Model after Instant Bleed-Down
Regression Residual Total df 5 26 31 SS 3050513 1058931 4109444 MS 610102.6 40728.13 F 14.98 Significance F 5.93E-07

For individual coefficient, it was found that four of five parameters contribute significantly to the late time behavior. Those critical parameters are formation pressure, annular conductivity, mud compressibility and initial gas concentration. Among these, initial gas concentration was the most critical one ( b 4 = 211.8 ) and mud compressibility

the least ( b 3 = -92.9 ). In this situation, regression models of pressure bleed-down and one-day buildup are enough for interpreting all five factors considered in SCP testing. Table 5. 8 Coefficient Estimates for Buildup after Instant Bleed-Down
Intercept pf k cm Fg Vgc Coefficients Standard Error 669.719 35.7 111.281 35.7 148.656 35.7 -92.906 35.7 211.844 -85.656 35.7 35.7 t Stat 18.772 3.119 4.167 -2.604 5.938 -2.401 P-value 1.22E-16 0.004 0.000 0.015 0.000 0.024

Buildup after Long Bleed-Down. ANOVA of fitted model (Table 5.8) shows that

the P-value (7.78E-7) is far less than 0.02, i.e., at least one factor contributes significantly to the model. Table 5. 9 ANOVA for One-day Pressure Buildup Model
Regression Residual Total df 5 26 31 SS 3126624 878046 4004670 MS 625325 33771 F 18.5 Significance F 7.78E-08

100

While checking the significance for individual coefficient, it was found that only three parameters contribute significantly to the one-day buildup regression model, since their P-values are less than 0.02 (Table 5.9). Among these, size of gas chamber was the most critical one ( b 5 = -197.1 ) and annular conductivity the least ( b 2 = 117.1 ). Table 5. 10 Coefficient Estimates for One-day Pressure Buildup
Intercept pf k cm Fg Vgc Coefficients Standard Error 467.9 32.5 48.3 32.5 117.1 32.5 -155.9 32.5 136.1 -197.1 32.5 32.5 t Stat 14.4 1.5 3.6 -4.8 4.2 -6.1 P-value 0.000 0.150 0.001 0.000 0.000 0.000

The result shows that effect of formation pressure on one-day pressure buildup is insignificant, i.e., no information about formation pressure can be withdrawn from the one-day buildup model. Continual buildup of two and three days was tested. P-value from ANOVA for two* day buildup (Eq.5.3) Dp 2 is 8.93E-7, showing that at least one factor significantly

contributed to the regression. From the estimation of individual factor (Table 5.10), it was found that formation pressure (P = 0.004) was one of three significant contributors and its effect ( b1 = -24.1 ) on 48-hour buildup is almost the same with that of annular conductivity ( b 2 = -25.8 ). Thus, two-day buildup is long enough to interpret all parameters concerned.
* Table 5. 11 Estimation of Individual Factor for Model of Dp 2

Intercept pf k cm Fg Vgc

Coefficients Standard Error 65.8 7.5 24.1 7.5 25.8 7.5 -7.2 7.5 -49.4 -17.3 7.5 7.5

t Stat 8.76 3.20 3.44 -0.96 -6.58 -2.31

P-value 0.000 0.004 0.002 0.347 0.000 0.029

Lower 95% Upper 95% 50.4 81.2 8.6 39.5 10.4 41.2 -22.6 8.2 -64.9 -32.7 -34.0 -1.9

101

Compared with two-day buildup, effect of formation pressure on three-day buildup (Eq.5.4) is still significant (Table 5.11) but effect of annular conductivity is diminishing (P = 0.018).
* Table 5. 12 Estimation of Individual Factor for Model of Dp 3

Intercept pf k cm Fg Vgc

Coefficients Standard Error 106.0 12.8 39.5 12.8 32.5 12.8 -6.3 12.8 -77.1 -23.7 12.8 12.8

t Stat 8.28 3.09 2.54 -0.49 -6.02 -1.85

P-value Lower 95% Upper 95% 0.000 79.7 132.3 0.005 13.2 65.8 0.018 6.2 58.8 0.628 -32.6 20.0 0.000 0.075 -103.4 -50.0 -50.8 2.6

5.3.5 Discussion of Sensitivity Analysis

Sensitivity analysis of bleed-down shows that final bled pressure is mainly affected by properties of mixture above the cement, such as gas chamber volume and initial gas concentration. Measuring gas chamber size could reduce the number of unknown parameters to one and eliminate the ambiguity of interpretation. Recording time interval should be reduced accordingly to detect discontinuity of pressure trend to obtain information about gas chamber. The quality of buildup records is dependent on the schedule of bleed-down used. One-day buildup after the instant bleed-down is sufficient to interpret all properties concerned. However, it is found that one-day buildup after the long bleed-down missed information about the formation. Further studies show that formation pressure is one of the controlling parameters for two-day pressure buildup. Since MMS recommended long bleed-down without liquid removal in SCP tests, one-day buildup is incomplete to understanding all properties affecting SCP. Therefore, the time limitation of pressure buildup should be extended to at least two days.

102

5.4 Field Validation of SCP Test Model

The mathematical model was verified with selected data from two wells, Well 19 and Well 25. Inputs for two wells are listed in Table 5.12. By matching SCP monitored records with calculated pressure curves, an effective method of analyzing SCP diagnostic test was developed. The match was conducted by changing values of five parameters controlling SCP. The matched results are listed in Table 5.13. The results provide the interpretation of SCP test and qualitative understanding of present SCP system. Table 5. 13 System Properties for Theoretical Calculation System Properties Unit Well 19 Inner diameter in 9.95 Annulus Outer diameter in 7.625 Length ft 598 Cement Porosity 0.01 Formation Depth ft 5447 Gas Specific gravity 0.71 Consistency index eq cp 285.64 Flow behavior index 0.8 Liquid Density ppg 11 Interface tension dynes/cm 67.5 Table 5. 14 Matched Results for Two Wells Parameters Unit Well 19 Well 25 Initial Lg ft 23 ft 1.2 ft Initial Fg 0.0025 0.0008 cm psi-1 6e-6 4e-6 psi-1 kA md-ft2 4.4 1.1 md-ft2 pf psi 4470 4455 psi
5.4.1 Validation with Data from Well 19

Well 25 9.95 7.625 597 0.01 5447 0.71 285.64 0.8 15 67.5

The schematic for well 25 (Table 5.12) was used for Well 19 because of missing data. The annulus between 7.625-in. production casing and the 10.75-in. intermediate casing in Well 19 has been bled for 12 minutes before the needle valve is closed and followed casing buildup lasts more than 24 hours. Essentially dry gas was bled from the 103

annulus. The hypothetical depth of gas source formation is at the intermediate casing shoe, which is 5447 ft from the wellhead (Table 5.12). The length of cement is 598 ft. The length of liquid column needs to be determined since the size of gas chamber is unknown. The match is done by visually minimizing the difference between recorded data and calculated data. To match the bleed-down history, two parameters, initial gas concentration and size of gas chamber, were mainly changed. In practice, the ambiguity of matching can be eliminated by measuring or collecting one of them. With matched bleed-down, matching the buildup history was continuously conducted by changing values of annular conductivity, mud compressibility and formation pressure. Five parameters obtained from match are listed in following table. Match plot is shown in Fig.5.13. The match is reasonably good. The number of unknown parameters can be reduced by collecting and sampling removed fluids, making the model a more feasible tool to analyze test data.
1600 kA = 4.4 md-ft2 pf = 4470 psi 1200

Casing Pressure (psia)

800

400 Theoretical Data Field Data 0 0 500 Time (min) 1000 1500

Fig.5. 13 Match of SCP bleed-down and buildup in Well 19

104

5.4.2 Validation with Data from Well 25

For this well, operators tried to inject heavier mud into the annulus between intermediate and production casing to reduce or eliminate SCP. Fig.5.14 is a schematic of Well 25, which had SCP problem in the intermediate (10 - in) casing. After pumping about 18,000 lbs of mud with density 15 ppg to 19 ppg into the intermediate casing, casing pressure bleed-down and buildup began. The hypothetical depth of gas formation was 5447 ft. And the length of cement was about 600 ft. Pressure has been bled for 14 seconds. 30 gal of 17.2 ppg mud was removed with gas.

565 1000 3760

Drive Pipe 26 Conductor Casing 20 94# H-40 Surface Casing 16 75# K-55

5447

Intermediate Casing 10 3/4 45.5# K-55 Production Casing 7 5/8 33.7# S-95

8635

Fig.5. 14 Schematic of Well 25 Since a clear usage of data had not been developed then, casing pressure buildup was recorded daily. Calculations using the model showed that those data missed important information on early buildup (Fig.5.15), since transition begins at about 2 hours after the bleed-down. Five matched parameters are list in Table 5.13. Results show that the length of gas chamber is shorter than previous case. This explains the instant bleed-down and removal of mud in this well. Even though there is much less gas in mud initially in this case 105

(0.0008) than in the previous case (0.025), instant bleed-down and additional hydrostatic pressure loss caused by mud removal draw more gas from cement flowing into mud than long bleed-down. Therefore, the early-time increase (393 psi) is more obvious than in previous example (183 psi) (Appendix F). But low annular conductivity and formation pressure leads to prolonged transition. After four days, the casing pressure increased less than 100 psi, comparing with that pressure increase about 180 psi in one day for Well 19.
800 kA = 1.1 md-ft2 pf = 4455 psi

Casing Pressure (psia)

600

400

200 Theoretical Data Field Data 0 0 2000 Time (min) 4000 6000

Fig.5. 15 Match of SCP bleed-down and buildup in Well 25

106

CHAPTER 6. SUMMARY AND CONCLUSIONS

As a result of the work presented on this dissertation, the following conclusion are drawn: 1. A databank including 26 wells from an oil field in GOM was built. The occurrence and magnitude of the SCP problem were addressed statistically. Similar trends were seen to those reported by MMS for the whole GOM. Thus, the SCP problem is widespread and independent from conditions of specific oilfield in the GOM. Also, the analysis method validated for one oilfield should work anywhere in the GOM. 2. Five patterns of SCP bleed-down and buildup were summarized from the databank. Those are instant and long bleed-down patterns, normal, s-shape and incomplete buildup patterns. 3. Current SCP diagnostic testing procedures and the MMS guidelines for the test are analyzed. From critical analysis of the procedures, it was concluded that there is a necessity to develop theoretical models to analyze diagnostic testing data. 4. Literatures about three controlling mechanisms of SCP, including invasion of gas into casing annuli, gas migration through cement and liquid column, and accumulation/discharge at the wellhead, were investigated. 5. Two basic of gas migration above the leaking cement was discovered: fast pass through the column of low viscosity Newtonian fluid, and, slow raising gas bubble swarms in a column of viscous, non-Newtonian mud. In this study, two mathematical models were developed to describe these scenarios.

107

6. In first model, gas flow in liquid column was described as fast gas migration through the column of low viscosity Newtonian fluid, assuming no gas being entrapped in liquid column. Effects of four factors on SCP are investigated using the model. It was found that: Early pressure buildup was controlled by mud compressibility, annular conductivity, and gas chamber; Late buildup was mostly influenced by formation pressure.

7. Casing pressure responses calculated by first model were used to compare with buildup records from two wells with SCP. The excellent match result demonstrated the feasibility of the model. 8. In second model, gas flow in liquid column was described as migration of gas bubble swarms in a column of viscous, non-Newtonian liquid, using a two-phase drift-flux model. It coupled numerical solutions of pressure and gas concentration distribution in liquid column with the analytical solution of gas flow in cement to calculate casing pressure during buildup and bleed-down. 9. Applications of the second model included: Understanding of typical SCP patterns and several SCP diagnostic test scenarios; Sensitivity analysis of five factors on pressure bleed-down and buildup; Field validations.

10. From sensitivity analysis, it was found that pressure responses in bleed-down and buildup were affected by different properties.

108

Pressure bleed-down is controlled by properties above the cement, such initial gas concentration in the liquid column and size of gas chamber.

One-day pressure buildup after instant bleed-down is enough to obtain information on both cement channeling and formation pressure.

After long bleed-down, longer buildup is necessary to obtain information about formation pressure. Therefore, 24 hours of buildup required by MMS is subject to debate.

11. By matching the calculated curves with field data, it is illustrated that the model is a potential for quantitative analysis of diagnostic testing in wells with SCP. 12. An improved program for modeling pressure bleed-down and buildup in SCP diagnostic test and analyzing testing results was developed. As the result of study, it is recommended that additional work should be focused on: 1. To improve the interpretation of SCP tests, it is suggested that records on pressure bleed-down be as accurate and complete as possible. Pressure history is definitely necessary. Besides that, collecting and sampling removed fluids (gas and liquid) is essential to understand the source of leakage and rheological properties of liquid. Current bleed-down is not an efficient procedure. Operators spend too much time to record the stabilization. Also, recording time interval should be reduced accordingly to detect pressure stabilization and discontinuity of pressure trend to obtain information about gas chamber. For instant bleeddown, records at every one to two second are necessary. 2. Detailed description of the procedure, such as the time when liquid beginning to flow through the needle valve, is very helpful to estimate gas chamber size and

109

eliminate the ambiguity of interpretation. No needle valve record has been found in previous field data. But the size of opening is essential to calculate flow rates. 3. For buildup, the MMS required time limitation (24 hours) seems to be arbitrary. Therefor, further research should be conducted to find a reasonable and feasible number. 4. In the mathematical model, liquid flow down to cement was not considered and only gas flow through the interface. In some cases, this assumption caused a pressure discrepancy at the interface when gas flow stopped. The effect of back flow should be investigated the future model. 5. Experiments should be performed to validate and improve current mathematical model developed. Also, some phenomena, such bubble generation and back flow, can be investigated by experiments.

110

REFERENCES
Achong, I., Revised Bean Performance Formula for Lake Maracaibo Wells, International Company Report, Shell Oil Co., Houston (October 1961). Al-Hussainy, R., Ramey, H. J., Jr. and Crawford, P. B., The Flow of Real Gases Through Porous Media, J. Pet. Tech., pp 624-636 (May 1966) Trans. AIME. Almehaideb, R. A, Aziz, K., and Pedrosa, O. A., A Reservoir/Wellbore Model for Multiphase Injection and Pressure Transient Analysis, paper SPE 17941 presented at the 1989 SPE Middle East Oil Technical Conference & Exhibition, Manama, Bahrain, March 11-14. Ansari, A. M. et al, A Comprehensive Mechanistic Model for Two-Phase Flow in Wellbores, SPEPF (May 1994) 143; Trans., AIME, 297. Ashford, F. E. and Pierce, P. E., Determining Multiphase Pressure Drops and Flow Capacities in Downhole Safety Valves, JPT (September 1975) 1145. Barnea, D., A Unified Model for Predicting Flow-Pattern Transition for the Whole Range of Pipe Inclinations, Intl. J. Multiphase Flow (1987) 13, 1. Baxendell, P. B., Bean Performance Lake Maracaibo Wells, International Company Report, Shell Oil Co., Houston (October 1967). Beggs, H. D., Gas Production Operations, OGCI publications, Tulsa, Oklahoma (1984). Bird, R. B., Stewart, W. E., and Lightfoot, E. N., Transport Phenomena John Wiley & Sons, New York City (1960). Blasius, H., Z. Math. Phys. (1908) 56, 1. Boehme, G., Strmung nicht-newtonscher Medien, Z. angew. Mathe. Und. Mech., 62 (1982) T3-T17. Bornea, D., and Taitel, Y., Flow Pattern Transitions in Two-Phase Flows, Encyclopedia of Fluid Mechanics, Vol. 6, 1986. Bourgoyne, A. T. Jr. et al., Applied Drilling Engineering, Textbook Series, SPE, Richardson, Texas (1991) 2. Bourgoyne, A. T. Jr., Scott, S. and Manowski, W., A Review of Sustained Casing Pressure (SCP) Occurring on The OCS, Paper presented at the LSU/MMS Well Control Workshop held on April 1, 1998. Brill, J. P. and Beggs, H. D., Two-Phase Flow in Pipes, U. of Tulsa, Tulsa, Oklahoma (1991).

111

Caetano, E. F., Upward Vertical Two-phase Flow Through an Annulus, PhD Dissertation, The University of Tulsa, Tulsa, Oklahoma, 1995, 226p. Caetano, E. F., Shoham, O. and Brill, J. P., Upward Vertical Two-Phase Flow Through an Annulus, Part I: Single-Phase Friction Factor, Taylor Bubble Rise Velocity and Flow Pattern Prediction, J. Energy Res. Tech. (March 1992) 114, 1. Cheremisinoff, N. P, Encyclopedia of fluid mechanics, Vol. 3,Houston: Gulf Pub. Co., Book Division, c1986. Colebrook, C. F., Turbulent Flow in Pipes with Particular Reference to the Transition Region Between the Smooth and Rough Pipe Laws, J. Inst. Civil Eng. (1939) 11, 133. Davies, R. M. and Taylor, G., The Mechanics of Large Bubbles Rising Through Extended Liquid And Through Liquid in Tubes, Proc., Royal Soc. London (1949) 200A, 375. Dodge, D. W. and Metzner, A. B., Turbulent Flow of Non-Newtonian Systems, AICHE J. (1959) 5, 189. Drew, T. B., Koo, E. C., and McAdams, W. H., Trans., AICHE (1930) 28, 56. Dukler, A. E. and Taitel, Y., Flow Regime Transitions for Vertical Upward Gas-Liquid Flow: A Preliminary Approach Through Physical Modeling Progress Report No. 1 NUREG-0162, January 1977. Duns, H. Jr., and Ros, N. C. J., Vertical Flow Of Gas And Liquid Mixtures From Boreholes, Proc. 6th World Petroleum Congress, Frankfurt (June 1963). Ebert, F., Strmung nicht-newtonscher Medien, Verlag Fr. Vieweg & Sohn, Braunschweig-Wiesbaden 1980. El-Saden, M. R., Heat Conduction in An Eccentrically Hollow, Infinitely Long Cylinder with Internal Heat Generation, J. Heat Transfer (1961) 83, 510. Fair, W. B. Jr., Pressure Buildup Analysis with Wellbore Phase Redistribution, SPE 8206 presented at the 64th Annual Fall Technical Conference and Exhibition, Las Vegas, NV, September 23-26, 1979. Forsynthe G. E. and Moler, C. B., Computer Solution of Linear Algebraic System, Prentice-Hall Inc. Englewood Cliffs, NJ, 1967. Forsythe, G. E. and Wasow, R. W., Finite Difference Methods for Partial Differential Equations, John Wiley & Sons Inc. New York, 1967. Fortunati, F., two-phase flow through wellhead chokes, paper SPE 3742 presented at the 1972 SPE European Spring Meeting, Amsterdam, 16-18 May.

112

Gilbert, W. E., Flowing and Gas-Lift Well Performance, Drill. & Prod. Prac. (1954) 126. Gould, T. L., Vertical Two-Phase Steam-Water Flow in Geothermal Wells, J. Petroleum Tech. 26, 833 (1974). Govier, G. W., and Aziz, K., the flow of complex mixtures in pipes, Van Nostrand Reinhold Co. (1972). Griffith, P. and Wallis, G. B., Two-Phase Slug Flow J. Heat Transfer (August 1961) 83, 307. Gunn, D. J. and Darling, C. W.W., Fluid Flow and Energy Losses in Non Circular Conduits, Trans., AICHE (1963) 41, 163. Haciislamoglu, M. and Langlinais, J., Non-Newtonian Flow in Eccentric Annuli, J. Energy Res. Tech. (June 1990) 112, 163. Hagedorn, A. R., and Brown, K. E., Experimental Study of Pressure Gradients Occurring during Continuous Two-Phase Flow in Small-Diameter Conduits, JPT (April 1965) 475-84. Harlow, F. H. and Amsden, A. A., A numerical Fluid Dynamics Calculation Method for All Flow speed, Journal of Computational Physics, Vol. 8, No. 2 (October 1971). Harmathy, T. Z., Velocity of Large Drops and Bubbles in Media of Infinite or Restrict Extent, AICHE J. (1960) 6, 281. Hasan, A. R. and Kabir, C. S. and Rahman, R., Predicting Liquid Gradient in A Pumping-well Annulus, SPEPE (February 1988) 113; Trans., AIME, 285. Hasan, A. R. and Kabir, C. S., A Mechanistic Approach to Understanding Wellbore Phase Redistribution, SPE 26483 presented at the 1993 SPE Annual Technical Conference and Exhibition, Houston, Texas, October 3-6. Hasan, A. R. and Kabir, C. S., Modeling Changing Storage During a Shut-in Test, SPE 24717 presented at the 1992 SPE Annual Technical Conference and Exhibition, Washington DC, October 4-7. Hasan, A. R. and Kabir, C. S., Two-Phase Flow in Vertical and Inclined Annuli, Intl. J. Multiphase Flow (1992) 18, 279. Hewitt, G. F. and Roberts, D. N., Studies Of Two-Phase Flow Patterns By Simultaneous X-ray and Flash Photography, United Kingdom Atomic Entergy Authority Report AERE-M 2159 (1969). Heyda, J. F., A Greens Function Solution for the Case of Laminar Incompressible Flow Between Non-Concentric Circular Cylinders, J. Franklin Inst. (January 1959) 267, 25. 113

http://www.ideal-aerosmith.com/flowcal.htm#Gases. Ishii, M., and Mishima, K., Study of Two-Fluid Model and Interfacial Area, NUREG/CR-1873, ANL-80-111 (December 1980). Kataoka, I. and Ishii, M., Mechanism and Correlation of Droplet Entrainment and Deposition in Annular Two-Phase Flow, ANL-82-44, NUREG/CR-2885, July 1982. Lahey, R. T., Jr., and Moody, F. J., The Thermal Hydraulics Of A Boiling Water Nuclear Reactor, American Nuclear Society, LaGrange Park, Illinois, 1977. Liles, D. R. and Reed, Wm. H., A Semi-Implicit Method for Two-Phase Fluid Dynamics, Journal of Computational Physics, Vol. 26, pp. 390 407 (1978). Mandhane, J. M., Gregory, G. A., and Aziz, K., A Flow Pattern Map for Gas-Liquid Flow in Horizontal Pipes, Int. J. Multiphase Flow, Vol. 1, pp. 537-553, 1974. Maurice, B. D., Malcolm N. G., and Pawel A. Nawrocki, Why Oilwells leak: Cement Behavior and Long-Term Consequences, SPE 64733 presented at the SPE International Oil and Gas Conference and Exhibition in China, Beijing, China, 7-10 November 2000. McCoy, J. N., Podio, A. L. and Huddleston, C. L., Acoustic Determination of Producing Bottomhole Pressure, SPEFE pp. 617-621 (September 1988). Metzner, A. B. and Reed, J. C., Flow of Non-Newtonian Fluids Correlation of The Laminar, Transition, and Turbulent-Flow Regions, AICHE J. (1955) 1, 434. Meunier, D., Wittmann, M. J., and Stewart, G., Interpretation of Pressure Buildup Test Using In-Situ Measurement of Afterflow, JPT (Jan. 1985) 143-52. Mishima, K. and Ishii, M., Flow Regime Transition Criteria Consistent with Two-Fluid Model for Vertical Two-Phase Flow, ANL-83-42, NUREG/CR-3338, April 1983. Myers, R. H. and Montegormery, D. C., Response Surface Methodology: Process and Product Optimization using Designed Experiments, New York: Wiley, pp700 (1995). Nguyen, D. L., Winter, E. R. F., and Greiner, M., Sonic Velocity in Two-Phase Systems Intl. J. Multiphase Flow (1981) 7, 311. Nikuradse, J., Compte Rendus (1840) 11, 961 and 1041; (1841) 12, 112. Nind, T. E. W., Principles of Oil Well Production, McGraw-Hill Book Co. Inc., New York City (1964). Omana, R. et al., Multiphase Flow Through Chokes, paper SPE 2682 presented at the 1969 SPE Annual Fall Meeting, Denver, Colorado 28 September 1 October, 1969. Orkszewski (1967), J., Predicting Two-Phase Pressure Drops In Vertical Pipes, JPT (June 1967) 829; Trans., AIME, 240. 114

Patankar, S. V., Numerical Hear Transfer and Fluid Flow, Hermosa Publishing Corp, 1980. Perkins, T. K., Critical and Subcritical Flow of Multiphase Mixtures Through Chokes, SPEDC (December 1993) 271. Pitzer, S. C., Rice, J. D., and Thomas, C. E., A Comparison of Theoretical Pressure Buildup Curves with Field Curves Obtained From Bottomhole Shut-in Tests, Trans, AIME (1959) 216, 416-19. Rader, D. W., Bourgoyne, A. T., and Ward, R. H., Factors Affecting Bubble-rise Velocity of Gas Kicks, JPT (May, 1975). Ros, N. C. J., An Analysis Of Critical Simultaneous Gas-Liquid Flow Through A Restriction and Its Application To Flowmetering, Appl. Sci. Res. (1960) 9. Rouhani, S. Z., and Sohal, M. S., Two-Phase Flow Patterns: A Review of Research Results, Progress in Nuclear Energy, Vol. 11, No. 3, pp. 219-259, 1983. Sachdeva, R. et al., Two-Phase Flow Through Chokes, paper SPE 15657 presented at the 1986 SPE Annual Technical Conference and Exhibition, New Orleans, 5-8 October. Sangani, A. S., and Acrivos, A., Int. J. Multiphase Flow, 9:181 (1983). Santos, Otto L. A. and Azar, J. J., A Study on Gas Migration in Stagnant NonNewtonian Fluids, paper SPE 39019 presented at the Fifth Latin American and Caribbean Petroleum Engineering Conference and Exhibition, Rio de Janeiro, Brazil, 30 August 3 September 1997. Sas-Jaworsky, A. II, Coil Tubing Operations and Services Part 4, World Oil (March 1992) 71. Scott, S. L. and Kouba, G.E., Advances in Slug Flow Characterization for Horizontal and Slightly Inclined Pipelines, paper SPE 20628 presented at the 1990 SPE Annual Technical Conference and Exhibition, New Orleans, 23-26 September. Snyder, W. A. and Goldstein, G. A., An Analysis of Fully Developed Laminar Flow in an Accentric, AICHE J. (1965) 11, 462. Somei, N., Mechanisms of Gas Migration after Cement Placement and Control of Sustained Casing Pressure, M. S. thesis, Louisiana State University, Baton Rouge, Louisiana, 1992, 154p. Stegemeier, G. L. and Matthews, C. S., A Study of Anomalous Pressure Build-up Behavior, Trans, AIME (1958) 213, 44-50. Sternling, V. C., Two Phase Flow Theory and Engineering Decision, award lecture presented at AICHE Annual Meeting (December 1965). 115

Szlias, A. P., Bobok, E., and Navratil, L., Determination of Turbulent Flow Pressure Loss For Non-Newtonian Oil Flow in Rough Pipes, Rheologica Acta (1981) 20, No. 5. Taitel, Y. and Dukler, A. E., A Model for Predicting Flow Regime Transitions in Horizontal and Near Horizontal Gas-Liquid Flow, AICHE, J., Vol. 22, No. 1, pp. 47-55 (1976). Taitel, Y., Barnea, D., and Dukler, A. E., Modeling Flow Pattern Transitions for Steady Upward Gas-Liquid Flow in Vertical Tubes, AICHE J., 26, 345 (1980). Walker, C. P., Determination of Fluid Level in Oil Wells by the Pressure-wave Echo Method, presented at the Los Angeles Meeting, October 1936, Transactions of AIME, 1936. Wallis, G. B., One Dimensional Two-Phase Flow, McGraw-Hill (1969). Winterfeld, P. H., Simulation of Pressure Buildup in a Multiphase Wellbore/Reservoir System, SPEFE (June 1989) 247-52. Xiao, J. J., Alhanati, F., and Reynolds, A. C., Modeling and Analyzing Pressure Buildup Data Affected by Phase Redistribution in the Wellbore, SPE Advanced Technology Series, Vol. 4, No. 1, pp. 28-37. Zuber, N and Hench, J., Steady State and Transition Void Fraction of Bubble Bubbling System and Their Operating Limits Part I: Steady State Operation, Report#62 GL100, General Electirc Co., Schenectady, New York (1962). Zuber, N. and Findlay, J. A., Average Volumetric Concentration in Two-Phase Systems, Trans, ASME, J. of Heat Transfer, Vol. 87, pp. 453-468 (November 1965).

116

APPENDIX A. DATA BANK AND SOFTWARE

Well # 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 NA NA Note: NA Not Applicable File Name MUA1.xls MUA2.xls MUA3.xls MUA4.xls MUA5.xls MUA6.xls MUA7.xls MUA8.xls MUA9.xls MUA10.xls MUA11.xls MUA12.xls MUA13.xls MUA15.xls MUA16.xls APTA19.xls APTA30.xls APTA31.xls APTL9.xls BPTB6.xls PTCA25C.xls PTCA7D.xls B7.xls HIA1.xls HIA2.xls HIA3.xls EarlyModel.xls SCPTestModel.xls

117

APPENDIX B. SCP BUILDUP IN WELLS WITH GAS-FREE LIQUID COLUMN ABOVE CEMENT
B.1. Derivations The schematic for this model is shown in Fig.3.1. The boundary and initial conditions are: Boundary Conditions:
p (0, t ) = Pf
q ( L, t ) = 0

(B. 1) (B. 2)

Initial Conditions:

p 2 ( z, t ) = p 2 f -

q 0 p scTzm i Z i 0.003164kTsc A

0 < z < Lc
Lc z < L

(B. 3) (B. 4)

p( z , t ) = 14.7 + 0.052 r m ( z - Lc )

In time step n, the pressure at cement top is:

pcn = ptn -1 + 0.052 r m Lnf-1 and flow rate is, q cn = 0.003164kTsc A 2 p f - p cn p scTLc m i Z i

(B. 5)

( )]
2

(B. 6)

From gas law, increase of gas moles at each time step and accumulative moles at gas chamber can be respectively obtained as
Dn n =

p sc qcn Dt Z i R 'Twb

(B. 7)
n

ntn = Dn k =
k =1

p
k =1

sc

qck Dt (B. 8)

Z i R 'Twb

118

Considering the expansion of gas and the compressibility of fluid, the following equations can be obtained:
ptn Vt n-1 + DVt n = ntn ZR 'Twh
n n -1 DV m = c mVm ( ptn - p tn-1 )

(B. 9) (B. 10)

and expansion of gas chamber is equal to the expansion of mud column, therefore,
n DV m = DV t n

(B. 11)

Substitute Eq.B.8 and Eq.B.10 into Eq.B.9, we can get:

Twh p sc q ck Dt
k =1 n -1 c mVm Twb n

(p )

n 2 t

n -1 n Vt - pt - c V n-1 p t m m
n -1

=0

(B. 12)

The root of this equation is pt at n-th time step:

n k 2 D 4 T p q t wh sc c n -1 n -1 V V 1 k =1 ptn = ptn -1 - t n -1 + ptn -1 - t n-1 + n -1 2 c mVm c mVm c mVm Twb

(B. 13)

B.2. Solution Diagram and Algorithm

The solution diagram can be see in Fig.B.1. At the beginning of each time step, using
Eq.B.5, pc is calculated according to the previous pt and keeps as constant. So the flow

rate of each time step is also constant, according to the steady-state flow assumption (Eq.B.6). In one time step, gas releases at cement top and immediately reaches at the gas chamber. Using the real gas law, the surface pressure pt at the end of this time step can be obtained (Eq.B.13). Therefore, in a step-wise manner, pressures at cement top and at casing head as functions of time can be computed.

119

Start Time n = 0 Use Eq. B.5, calculate pressure pc

Use Eq. B.6, calculate flow rate qc

Use Eq. B.13, calculate pt n=n+1

tn >tmax
Y

Stop

Fig.B. 1 Solution diagram for gas-free mud column model

B.3. Nomenclature
A= p 2 ( D12 - D2 ) = wellbore area, sq ft 4

c m = mud compressibility, psi-1

D1 = outer diameter, ft D2 = inner diameter, ft

k = cement permeability, md

Lc = length of cement column, ft Lt = length of gas chamber, ft

L f = length of mud column, ft

n = moles pc = pressure on the top of the cement, psia

p f = reservoir pressure (constant), psia

120

pt = surface pressure, psia qc = flow rate on the top of the cement, SCF/D
R = 10.73 = gas constant
T = reservoir condition temperature, o R Twb =

1 (T + Twh ) = average wellbore temperature, o R 2

Twh = wellhead temperature, o R (usually 520 o R )

Dt = time step, day

tmax = time to stop, day Vm = volume of mud column, cu ft Vt = volume of gas chamber, cu ft
Z = gas-law deviation factor, dimensionless
z = distance from the gas-source formation, ft

m g = gas viscosity, cp
r m = density of mud in wellbore, ppg

121

APPENDIX C. SOLUTION OF THE DIFFERENTIAL EQUATION FOR THE LINEAR GAS FLOW
C.1 The Continuity Equation Considering the three-dimensional case, we choose an arbitrary rectangular volume as shown in Fig. C.1.
Uz+(Uz) Ux

Uy y Ux+(Ux)

z x

Uy+(Uy) Z

Y Uz X

Fig.C. 1 Volume element in three dimensions The volumetric components of flow into the element in the x, y and z direction are denoted by u x , u y and u z respectively. These are volumetric flow rates per unit of crosssectional area. Thus, the mass flow rate into the element in x-direction is ru x DyDz . The mass flow rate in the x-direction out of the element is DyDz[ru x + D (ru x )]. And D( ru x ) is the change in mass flux that occurs within the element. The net flow rate in the xdirection (amount-in less amount-out) is - DyDzD (ru x ) . Similar expressions can be written for the y and z directions. Assuming no mass is generated or lost in the element, the amount of net mass change in the element in a time increment t can be expressed as
- Dt D (ru x )DyDz + D (ru y )DzDx + D (ru z )DxDy = fr DxDyDz |t + Dt -fr DxDyDz |t

122

Dividing the equation by xyzt yields

D( ru x ) D (ru y ) D ( ru z ) ( rf )t + Dt - (rf )t - + + = Dt Dy Dz Dx
Proceeding to the limit as x, y, z and t approach zero gives (ru x ) + (ru y ) + (ru z ) = - (rf ) x y z t This is the continuity equation for flow of a fluid in a porous medium. Assuming no flow in x and y direction, we can simplify it into one dimension (z-direction) as
(ru z ) = - (rf ) z t

(C. 1)

To derive differential equation for fluid flow in a porous medium, Darcys law must be combined with the continuity equations. For the one-dimensional case, the differential equation is k z r p + rg = (fr ) z m z t (C. 2)

This is a general form for the combination of the continuity equation and Darcys law. C.1.1 Single-Phase Gas Flow To develop a solution for gas flow, two additional equations are required: the realgas law:
r= M p RT Z

(C. 3)

and the equation that describes how the gas isothermal compressibility varies with pressure:

123

cg =

1 1 dZ p Z dP

(C. 4)

If we neglect gravity, those three equations can be combined to yield p p fct p p = kZ t z mZ z where
ct = c g + c f

(C. 5)

Al-Hussainy, Ramey, and Crawford (1966) introduced a transformation of variables to obtain a solution for the differential equation. The transformation involves the real gas pseudopressure, m(p), which is defined as

m( p ) = 2

p dp mZ pR

(C. 6)

where p R is a reference pressure, usually chosen to be 14.7 psia, from which the function is evaluated. Since m and Z are only functions of pressure for a given reservoir system, which were assumed to be isothermal, the differential equation can be differentiated and the chain rule of differentiation applied to obtain the following relationships m( p ) 2 p = p mZ m( p ) m( p ) p = z p z m( p ) m( p ) p = t p t Substituting Eq.(C.7) into Eq.(C.8) and (C.9) yields p mZ m( p ) = z 2 p z 124 (C. 10) (C. 7)

(C. 8)

(C. 9)

p mZ m( p ) = t 2 p t Combining Eq.(C.10), (C.11), and (C.5) yields 2 m( p ) fm ct m( p ) = k t z 2

(C. 11)

(C. 12)

This equation is still a nonlinear differential equation because of the dependence of the m and ct on pressure or the real gas pseudopressure. Thus, there is no analytical solution for the above equation. Al-Hussainy and Ramey (1966) proposed that m and ct could be evaluated at the initial pressure. So the differential equation can be linearized as 2 m( p ) f (mct )i m( p ) = k t z 2 C.2 The Boundary and Initial Conditions The boundary and initial conditions may be stated as follows: (C. 13)

a ) p = pc at t = 0 for all z (except z = 0 ) b ) p = p f at z = L for all t kA p c) =q for t > 0 m z z = 0

(C. 14)

Condition (a) is merely the initial condition that, before producing, the pressure every where within the drainage volume is equal to the initial equilibrium pressure, pc. But it is equal to formation pressure pf at lower end. Condition (b) ensures that the pressure at the lower boundary is not affected by the pressure disturbance at the surface, and vice versa. Condition (c) ensures constant terminal rate at the upper boundary. To solve the differential equation of pseudopressure, those conditions need to be changed in terms of pseudopressure. By definition, they are

125

a ) m( p ) = m( p c ) at t = 0 for all z (except z = 0 ) b ) m( p ) = m( p f ) at z = 0 for all t m( p ) - 2 pq c) = for t > 0 z z = L kAZ

The Constant Terminal Rate Solution

(C. 15)

Under the above conditions, Eq. (C.13) can be solved by the method of separation of variables. In order for separation of variables to apply, the boundary conditions (BCs) must be linear and homogeneous. C.3 Transforming Nonhomogeneous BCs to Homogeneous Ones To change those nonzero BSs to homogeneous ones, a solution of the following form is seeked.

v( z , t ) = m( z , t ) - u ( z , t )

(C. 16)

where u(z, t) are chosen to satisfy the BCs of the problem. Substituting u(z, t) into the BCs
u (0, t ) = m( p f ) u 2 pq =kAZ z z = L

(C. 17)

Solving for u(z, t), we get

u (z , t ) = -

2 pq z + m( p f kAZ

(C. 18)

So, on substituting this into the original problem (Eq.C.13 and C.15), transformed problem in v(z, t) is obtained.

126

 2 v f (mct )i v PDE = k t z 2 v(0, t ) = 0 BCs v =0 z z = L 2 pq IC v( z ,0 ) = z kAZ

(C. 19)

New problem with homogeneous BCs can now be solved by separation of variables.
C.4 Solution for Transformed Problem

Substituting v(z,t) = Z(z)T(t) into the partial differential equation (PDE) gives
Z (z )T ' (t ) = a 2 Z '' (z )T (t )

where
a2 = k f (mct )i

Dividing each side by a 2 Z ( z )T (t ) , we have

T ' (t ) Z '' (z ) = a 2T (t ) Z (z )

and obtain separated variables. For z and t are independent of each other, each side must be a fixed constant (say k); hence, it can be written as
T ' (t ) Z '' ( z ) = =k a 2T (t ) Z (z )

or
T ' - a 2 kT = 0 Z '' - kZ = 0

Now solving each of these two ordinary differential equations (ODEs), multiply them together to get a solution to the PDE. To let the T(t) factor go to zero as t , the 127

separation constant k must be set to be negative. Usually it is done by setting k = -l2 . Thus, the two ODEs can be written as
T ' + a 2 l2 T = 0 Z '' + l 2 Z = 0

Both equations are standard ODEs and have solution

T (t ) = Ae - a l t
2 2

Z ( z ) = A sin (lz ) + B cos(lz )

and hence all functions

v(z , t ) = e - l a t [ A sin (lz ) + B cos(lz )]
2 2

(C. 20)

will satisfy the PDE vt = a 2 v zz . To choose a certain solution for this problem, Eq.(C.20) should satisfy the boundary conditions v(0, t ) = 0 v z (L , t ) = 0 Substituting Eq.(C.20) into these BCs, we get v(0, t ) = Be - l a t = 0 B = 0 2 2 v z (L, t ) = Ale -l a t sin (lL ) = 0 sin (lL ) = 0
2 2

This last BC restricts the separation constant l to be

l=
2n + 1 p 2L n = 0, 1, 2, ...

Now having found an infinite number of functions

2n + 1 2 2n + 1 v(z , t ) = An sin pz exp - pa t n = 0, 1, 2, ... 2 L 2L

(C. 21)

each one satisfying the PDE and BCs.

128

The last step is to add the fundamental solutions

2n + 1 2 2n + 1 v(z , t ) = An sin pz exp - pa t 2 L 2L n =0

in such a way that the initial condition (IC)

v ( z ,0 ) = 2 pq z kAZ

is satisfied. Substituting the sum into the IC gives

2 pq 2n - 1 z = An sin pz kAZ 2L n =1

(C. 22)

2n + 1 Multiplying each side of this equation by sin pz and integrating from 0 to L, 2L

2 pq L 2n - 1 2n - 1 z sin pz dz = An sin 2 pz dz = An 2 kAZ 0 2L 2L 0 Solving for An gives

An = (- 1)
n

pqL (2n + 1) p kAZ 16

2 2

(C. 23)

So the solution is 16 pqL v( z , t ) = kAZp 2

2n + 1 2 1 2n + 1 (- 1) sin pz exp - pa t (2n + 1)2 2 L n=0 2L

(C. 24)

C.5 Solution for Original Problem

Now the solution for original problem is

p 2 z 16 pqL 8 L m( z , t ) = 2 2 2n + 1 1 kAZp 2n + 1 n (- 1) pz exp - pa t sin 2 (2n + 1) 2 L 2L n=0 + m( p f )

(C. 25)

129

At the well (z = L), the pseudopressure drop is

16 pqL Dm ( L , t ) = kAZp 2 where
Dm ( L , t ) = m ( p f ) - m ( L , t )

2n + 1 2 1 p 2 - exp - pa t 2 2L 8 n =0 (2n + 1)

(C. 26)

C.6 Multi-Rate Solution

Mathematically, the superposition theorem states that any sum of individual solutions of a second order linear differential equation is also a solution of the equation. Consider the case of well producing at a series of constant rates for the different periods show in Fig. C.2.
q1 q2 q4 Rate ... t1 Pi Pwf ... t1 t2 t3 t4 tn time t2 t3 t4 tn time q3 qn

Fig.C. 2 Production history of a well To determine the well pressure after a total flow time tn when the current rate is qn, the superposition theorem is applied to determine a composite solution of Eq.(C.26) in terms of

130

q1 Acting for time t n + (q 2 - q1 ) Acting for time (t n - t1 ) + (q3 - q 2 ) Acting for time (t n - t 2 ) ... + (q 2 - q n -1 ) Acting for time (t n - t n -1 ) That is, a solution is obtained for the initial rate q1, acting over the entire period tn. At time t1 a new well is opened to flow at precisely the same location as the original well at (q2 - q1) so that the net rate after t1 is q2. At time t2 a third well is opened at the same location with rate (q3 q2) which changes the rate to q3 after time t2 etc. The composition solution of Eq.(C.13) for this variable rate case can then be formed by adding individual constant terminal rate solutions, Eq.(C.26), for the rate-time sequence specified above, i.e. m( p f ) - m( p wf ) = (q1 - 0)M (t n - 0) + (q 2 - q1 )M (t n - t1 ) M + (q n - q n -1 )M (t n - t n -1 ) where m( p wf

) is the specific value of the well flowing pseudopressure corresponding to

the total time tn, which may occur at any time during the n-th period of constant flow, when the rate is qn. And M (t ) = (2n + 1)p 16qmL 1 p 2 - exp - 2 2 kAp 2L 8 n = 0 (2n + 1) a
2

The summation can be expressed as m( p ) - m( p ) = Dq M (t

f wf j j =1 n

- t j -1 )

(C. 27)

where Dq j = q j - q j -1

131

C.7 Transiation from Pseudopressure to Pressure

All the parameters in the integrand of Eq.(C.6) are themselves functions of pressure and can be obtained directly from PVT analysis of the gas at reservoir temperature or, knowing only the gas gravity, from standard correlation of m and Z, again at reservoir temperature. Using a simple graphical method for numerical integration (trapezoidal rule), a table of values of m(p) can be generated as a function of the actual pressures. Having once obtained this relationship, the resulting table or plot should be preserved since it will be relevant for the entire life of the reservoir. Once the table of p vs. m(p) is made, any interpolation formula can be used to convert from real to pseudo pressures and vice versa.

132

APPENDIX D. MANUAL OF SOFTWARE: SCPTESTMODEL

D.1 General Guide The software consists of eight spreadsheets and their names are self-explanatory. The input begins with sheet CasingSchematic (Fig.D.1). Usually, cells with green background are for data input and unit conversions or simple calculations are automatically carried in cells filled with brown color. Click button Next Sheet to proceed to next sheet. To convert those inputs into the model, click the button Save. D.2 Sheets for Input Sheet: CasingSchematic. Properties for casing are entered in this sheet. There is also a schematic of casing showing where those data are located. Geometric information of casing and open-hole, such as depth, diameters need to be entered under the category Well. Area and volume are calculated automatically. Inputs under the category Cement are used to calculate the length of cement. If length of cement is known, it can be directly entered.

Fig.D. 1 Sheet for casing schematic Sheet: BasicInput. Information about gas and mud, annulus, cement and formation are entered in this sheet (Fig.D.2). Only green cells need to be entered. Clicking button 133

Save to save those inputs into the model. Then click Next Sheet to go to sheet BleedoffInput.

Fig.D. 2 Sheet for basic inputs Sheet: BleedOffInput. Input of bleed-down could be entered in this sheet (Fig.D.3). Parameters needed to change to obtain the match are entered under Input for Simulation. Gas chamber Volume is in standard condition. Average Fg is the initial gas concentration in liquid column and the pressure gradient is calculated automatically.

Fig.D. 3 Sheet for bleed-down inputs 134

Opening number of needle valve is used to determine gas flow characteristic cv and its default data is listed in Sheet Defaults(Fig.D.4). This data is used to calculate gas flow rate through the needle valve. If gas flow rates are known, they can be entered directly. If liquid was removed during bleed-down, its accumulative volume can be entered and its flow rate is calculated in the model.

Fig.D. 4 Default values of gas flow characteristic cv After saving input data, click button Simulate (Fig.D.3) to calculate theoretical pressure curve. And next sheet BOSimResults (Fig.D.5) will automatically show the results graphically and tabularly. Details of this sheet will be discussed in the following section. If the result is not satisfactory, click button Previous Sheet and change volume of gas chamber or initial gas concentration and click Simulate to check the result again. If the matching is good, click button Next Sheet to go to sheet BuildupInput.

135

Fig.D. 5 Simulation results of pressure bleed-down Sheet: BuildupInput. Buildup records are entered in this sheet (Fig.D.6). Click button Save to convert and save the information for model. Clicking button Simulation automatically leads to the sheet BUSimResults, which will be discussed in the following section.

Fig.D. 6 Input for pressure buildup

136

D.3 Sheets for Simulation Results Sheet: BOSimResults. As shown in Fig.D.5, simulation results for bleed-down are listed. At the bottom of the table, a brief summary lists the surface pressure, pressure above and below the mud-cement interface, and gas flow rate at the interface (Fig.D.6).

Fig.D. 7 Summary of bleed-down simulation Sheet: BUSimResults. The results of buildup are listed in tables (Fig.D.8) in the same way with sheet BOSimResults(Fig.D.5) without the chart.

Fig.D. 8 Simulation results of pressure buildup

137

Chart: SCPTestChart. Clicking button Next Chart on sheet BUSimResults brings out the chart, which shows the result graphically (Fig.D.9).
1600 900 800 700 600 1000 500 800 400 600 300 400 200 100 0 1600 Interface Gas Rate (SCF/D)

1400

1200 Casing Pressure (psia)

200 0 0 200 400 600 800 Time (min) Surface Pres BleedOff Data Buildup Data IF Rate 1000 1200 1400

Fig.D. 9 Simulation result chart If the result is not satisfying, change the information in sheet BasicInput(Fig.D.2), such as mud compressibility, formation pressure or annular conductivity. Repeat the simulation until the match is acceptable. After changing the basic inputs, matching for bleed-down may need to repeat.

138

APPENDIX E. ABSTRACT OF PAPER SPE 67194

Problem of sustained casing pressure (SCP) is widespread in the Gulf of Mexico. Recently, over 12,000 wells have been reported with casing pressure that can not be permanently bled off through needle valves at their wellheads. In some cases the pressure can reach dangerously high values. The US Minerals Management Service (MMS) regulations 30 CFR 250.517 require remedial operation on a well if any of its casing string has severe SCP problem. MMS has also developed guidelines to tolerate small values of SCP a departure from 30 CFR 250.517. However, wells with approved departure must be frequently tested so that severity of SCP could be monitored and controlled. Presently, testing of SCP is mostly qualitative and limited to arbitrary criteria for casing pressure buildup. Such information is insufficient for operators to quantitatively analyze SCP problem and prevent potential risks. Thus there is a need for improved analysis that could provide information on the well parameters causing gas migration and SCP. In this paper, a mathematical model for diagnosis SCP from bleed-off/buildup has been developed. The model simulates gas migration in the cement column and its accumulation at the well head. The paper presents validation of the model with field data. The validation reveals a correlation between the pressure buildup-stabilization pattern and well parameters; cement conductivity controls the pattern of pressure buildup, while the gas-source formation pressure controls the pressure stabilization. It is also shows that cement conductivity, gas-source formation depth and pressure become critical parameters for remedial designs.

139

Also presented in the paper are two examples of using this model to analyze actual SCP field tests. The analysis gives acceptable estimates of the gas-source formation depth and pressure, cement conductivity, and expected maximum casing pressure value.

140

APPENDIX F. PRESSURE HISTORY FOR WELL 19 & 25

Time min 0 2 4 6 8 10 12 14 34 64 94 124 154 184 214 244 274 304 334 364 394 424 454 484 514 544 574 604 634 664 694 724 754 784 814 844 874 904 934 964 994 1024 1054 1084 1114 1144 1174 1204 1234 1264 1294 1324 1354 1384 1414 1444 1474 Well 19 Pressure Recorded Theoretical psia 1354 1354 1322 1148 1104 1013 872 926 777 866 767 818 764 775 777 775 837 813 882 869 914 913 941 927 960 932 977 937 986 941 997 946 1002 950 1007 955 1011 959 1015 963 1020 967 1023 972 1026 976 1030 980 1034 984 1038 988 1043 992 1047 996 1051 1000 1055 1004 1055 1008 1054 1012 1058 1016 1060 1019 1063 1023 1065 1027 1068 1031 1071 1034 1073 1038 1075 1041 1077 1045 1080 1048 1080 1052 1083 1055 1085 1059 1088 1062 1090 1066 1093 1069 1095 1072 1097 1076 1100 1079 1103 1082 1104 1085 1107 1088 1110 1092 1113 1095 1116 1098
Well 25 Field Data Theoretical Data Time Pressure Time Pressure min psia min psia 0.2 55 0.0 695 840.2 495 0.2 55 2340.2 495 27.9 128 3780.2 535 54.8 229 5220.2 535 81.7 328 108.7 406 136.7 430 163.6 435 190.4 436 217.3 437 244.1 438 272.2 438 299.0 439 325.9 439 352.7 440 380.7 440 407.6 441 434.4 442 461.3 443 488.1 444 516.2 445 543.0 446 569.9 447 596.7 448 623.6 449 651.6 449 678.5 450 705.3 451 732.2 452 759.0 453 787.1 453 813.9 454 840.2 455 889.3 456 937.2 457 986.2 458 1034.1 459 1083.1 460 1131.0 461 1180.0 463 1227.9 464 1275.8 465 1324.9 466 1372.7 467 1421.8 468 1469.7 470 1518.7 471 1566.6 472 1614.5 473 1663.5 474 1711.4 475 1760.5 476 1808.4 477 1857.4 478 1905.3 479 1953.2 480 2002.3 481 2050.1 482 2099.2 484 2147.1 485 2196.2 486 2244.0 487 2291.9 488 2387.0 490 2433.7 490 2480.4 491 2527.1 492 2572.7 493 2619.4 494 2666.2 495

141

VITA
Rong Xu was born in Ulumuqi, Xinjiang, Peoples Republic of China. She obtained a Bachelor of Science in reservoir engineering, Petroleum University (East China), Dongying, China, 1993. She obtained her Master of Science degree in petroleum engineering, Petroleum University in Beijing, Beijing, China, 1996. Research work was on developing analytical testing models for horizontal and slanted well test. She worked and studied at Research Institute of Petroleum Exploration and Development (RIPED), Beijing, China from 1996 to 1998. She then entered the Craft & Hawkins Department of Petroleum Engineering at Louisiana State University in the spring 1999 to work towards a doctoral degree in petroleum engineering in December 2002.

142